The share package \AREP provides an infrastructure and high level functions to do efficient calculations in constructive representation theory. By the term ``constructive" we mean that group representations are constructed and manipulated up to equality -- not only up to equivalence as it is done by using characters. Hence you can think of it as working with matrix representations, but in a very efficient way using the special structure of the matrices occuring in representation theory of finite groups. The package is named after its most important class ARep (see AReps) ( Abstract Representations) \footnote A note on the name: We have chosen ``abstract" because we manipulate expressions for representations, not constants. However, ``concrete" would also be right because the representations are given with respect to a fixed basis of the underlying vector space. The name ARep is thus, for historical reasons, somewhat misleading. implementing this idea.
A striking application of constructive representation theory is the decomposition of matrices representing discrete signal transforms into a product of highly structured sparse matrices (realized in Matrix Decomposition). This decomposition can be viewed as a fast algorithm for the signal transform. Another application is the construction of fast Fourier transforms for solvable groups (realized in DecompositionMonRep). The package has evolved out of this area of application into a more general tool.
The package \AREP consists of the following parts:
• Monomial Matrices: A monomial matrix is matrix containing exactly one non-zero entry in every row and column. Hence storing and computing with monomial matrices can be done efficiently. This is realized in the class Mon, Sections Mons -- CharPolyCyclesMon.
• Structured Matrices: The class AMat, Sections AMats -- UpperBoundLinearComplexityAMat, is created to represent and calculate with structured matrices, like e.g. 2.(A⊕ B)C⊗ D. E2, where A, B, C, D, E are matrices of compatible size and characteristic.
• Group Representations: The class ARep, Sections AReps -- DecompositionMonRep, is created to represent and manipulate structured representations up to equality, like e.g. (φ\uparrowT G)M⊗ψ. Special care is taken of monomial representations.
• Symmetry of Matrices: In Sections Symmetry of Matrices -- PermIrredSymmetry functions are provided to compute certain kinds of symmetry of a given matrix. Symmetry allows to describe structure contained in a matrix.
• Discrete Signal Transforms: Sections Discrete Signal Transforms -- InverseRationalizedHaarTransform describe functions to construct many well-known discrete signal transforms.
• Matrix Decomposition: Sections Matrix Decomposition -- MatrixDecompositionByMonMonSymmetry describe functions to decompose a discrete signal transform into a product of highly structured sparse matrices.
• Tools for Complex Numbers, Matrices and Permutations: Sections Complex Numbers -- TensorProductPerm describe useful tools for the computation with complex numbers, matrices and permutations.
All functions described are written entirely in the
GAP3 language. The functions for the computation of
the symmetry of a matrix (see Symmetry of Matrices)
may use the external C program desauto
written by J. Leon
and contained in the share package \sf GUAVA.
However, the use of this program is optional and will only
influence the speed and not the executability of the functions.
The package \AREP was created in the framework of our theses where the background of constructive representation theory (see Pue98) and searching for symmetry of matrices (see Egn97) can be found.
After having started GAP3 the \AREP package needs to be loaded. This is done by typing:
\renewcommand\baselinestretch0.8
gap> RequirePackage("arep"); ___ ___ ___ ___ | | | | | | | Version 1.0, 16 Mar 1998 |___| |___| |___ |___| | | | \ | | by Sebastian Egner | | | \ |___ | Markus Pueschel Abstract REPresentations
\renewcommand\baselinestretch1
If \AREP isn't already in memory it is loaded and its
banner is displayed. If you are a frequent user of \AREP
you might consider putting this line into your .gaprc
file.
The class Mon is created to represent and calculate efficiently with monomial matrices. A monomial matrix is a matrix which contains exactly one non-zero entry in every row and every column. Hence monomial matrices are always invertible and a generalization of permutation matrices. The elements of the class Mon are called ``mons". A mon m is a record with at least the following fields.
isMon
& := & true
isGroupElement
& := & true
domain
& := & GroupElements
operations
& := & MonOps
char
& : & characteristic of the base fieldperm
& : & a permutationdiag
& : & a list of non-zero field elements
\endtabular
The MonOps class is derived from the GroupElementOps class, so that groups of mons can be constructed. The monomial matrix represented by a mon m is given by
[δip j| i,j∈{1,...,Length( m.diag ) }].
ApplyFunc(DiagonalMat, m.diag ) ,
|
m.perm
and δkl denotes the
Kronecker symbol (δkl = 1 if k = l and = 0 else).
Mons are
created using the function Mon
. The following sections describe
functions used for the calculation with mons.
Some remarks on the design of Mon: Mons cannot be mixed
with GAP3-matrices (which are just lists of lists of field
elements); use MonMat
(MonMat) and MatMon
(MatMon) to convert
explicitly. Mons are lightweighted, e.g. only the characteristic of
the base field is stored. Mons are group elements but there are no
efficient functions implemented to compute with mon groups. You should
think of mons as being a similar thing as integers or permutations:
They are just fundamental objects to work with.
The functions concerning mons are implemented in the file
"arep/lib/mon.g"
.
m1 = m2
m1 <> m2
The equality operator =
evaluates to true
if the mons m1
and m2 are equal and to false
otherwise. The inequality operator
<>
evaluates to true
if the mons m1 and m2 are not equal
and to false
otherwise.
Two mons are equal iff they define the same monomial matrix. Note that the monomial matrix being represented has a certain size. The sizes must agree, too.
m1 < m2
m1 <= m2
m1 >= m2
m1 > m2
The operators <
, <=
, >=
, and >
evaluate to true
if the
mon m1 is strictly less than, less than or equal to, greater than or
equal to, and strictly greater than the mon m2.
The ordering of mons m is defined via the ordering
of the pairs [m.perm, m.diag]
.
74.4 Basic Operations for Mons
The MonOps class is derived from the GroupElementsOps class.
m1 * m2
m1 / m2
The operators *
and /
evaluate to the product and quotient of the
two mons m1 and m2. The product is defined via the product of
the corresponding (monomial) matrices. Of course the mons must be of
equal size and characteristic otherwise an error is signaled.
m1 ^ m2
The operator ^
evaluates to the conjugate
<m2>-1* m_1* m_2 of
m1 under m2 for two mons m1 and m2.
The mons must be of equal size and characteristic
otherwise an error is signaled.
m ^ i
The powering operator ^
returns the i-th power of the mon
m and the integer i.
Comm( m1, m2 )
Comm
returns the commutator
<m1>-1* m_2-1* m_1*m_2 of two
mons m1 and m2. The operands must be of equal size
and characteristic otherwise an error is signaled.
LeftQuotient( m1, m2 )
LeftQuotient
returns the left quotient <m1>-1*m_2 of two
mons m1 and m2. The operands must be of equal size
and characteristic otherwise an error is signaled.
Mon( p, D )
Let p be a permutation and D a list of field elements ≠ 0 of
the same characteristic. Mon
returns a mon representing the
monomial matrix given by
[δip j| i,j∈{1,...,Length( D )
}].
ApplyFunc(DiagonalMat, D )
, where δkl denotes the
Kronecker symbol. The function will signal an error if
the length of D is less than the largest moved point of p.
gap> Mon( (1,2), [1, 2, 3] ); Mon( (1,2), [ 1, 2, 3 ] ) gap> Mon( (1,3,4), [Z(3)^0, Z(3)^2, Z(3), Z(9)]); Mon( (1,3,4), [ Z(3)^0, Z(3)^0, Z(3), Z(3^2) ] )
Mon( D, p )
Mon
returns a mon representing the monomial matrix given by
ApplyFunc(DiagonalMat, D )
.
[δip j| i,j∈{1,...,Length( D )
}],
where δkl denotes the Kronecker symbol.
Note that in the output the diagonal is commuted to the right side,
but it still represents the same monomial matrix.
gap> Mon( [1,2,3], (1,2) ); Mon( (1,2), [ 2, 1, 3 ] ) gap> Mon( [Z(3)^0, Z(3)^2, Z(3), Z(9)], (1,3,4) ); Mon( (1,3,4), [ Z(3^2), Z(3)^0, Z(3)^0, Z(3) ] )
Mon( D )
Mon
returns a mon representing the (monomial) diagonal matrix
given by the list D.
gap> Mon( [1, 2, 3, 4] ); Mon( [ 1, 2, 3, 4 ] )
Mon( p, d )
Mon( p, d, char )
Mon( p, d, field )
Let p be a permutation and d a positive integer. Mon
returns a mon representing the (d×d) permutation matrix
corresponding to p using the convention
[δip j| i,j∈{1,...,d}], where
δkl denotes the Kronecker symbol. As optional parameter
a characteristic char or a field can be supplied.
The default characteristic is zero. The function will signal
an error if the degree d is less than the largest moved point of p.
gap> Mon( (1,2), 3 ); Mon( (1,2), 3 ) gap> Mon( (1,2,3), 3, 5 ); Mon( (1,2,3), 3, GF(5) )
Mon( m )
Let m a mon. Mon
returns m.
gap> Mon( Mon( (1,2), [1, 2, 3] ) ); Mon( (1,2), [ 1, 2, 3 ] )
IsMon( obj )
IsMon
returns true
if obj, which may be an object of
arbitrary type, is a mon, and false
otherwise. The function
will signal an error if obj is an unbound variable.
gap> IsMon( Mon( (1,2), [1, 2, 3] ) ); true gap> IsMon( (1,2) ); false
IsPermMon( m )
IsPermMon
returns true
if the mon m represents a
permutation matrix and false
otherwise.
gap> IsPermMon( Mon( (1,2), [1, 2, 3] ) ); false gap> IsPermMon( Mon( (1,2), 2) ); true
IsDiagMon( m )
IsDiagMon
returns true
if the mon m represents a diagonal matrix
and false
otherwise.
gap> IsDiagMon( Mon( (1,2), 2) ); false gap> IsDiagMon( Mon( [1, 2, 3, 4] ) ); true
PermMon( m )
PermMon
converts the mon m to a permutation if possible and
returns false
otherwise.
gap> PermMon( Mon( (1,2), 5) ); (1,2) gap> PermMon( Mon( [1,2] ) ); false
MatMon( m )
MatMon
converts the mon m to a matrix (i.e. a list of lists
of field elements).
gap> MatMon( Mon( (1,2), [1, 2, 3] ) ); [ [ 0, 2, 0 ], [ 1, 0, 0 ], [ 0, 0, 3 ] ] gap> MatMon( Mon( (1,2), 3) ); [ [ 0, 1, 0 ], [ 1, 0, 0 ], [ 0, 0, 1 ] ]
MonMat( M )
MonMat
converts the matrix M to a mon if possible and returns
false
otherwise.
gap> MonMat( [ [ 0, 1, 0 ], [ 1, 0, 0 ], [ 0, 0, 1 ] ] ); Mon( (1,2), 3 ) gap> MonMat( [ [ 0, 1, 0 ], [ E(3), 0, 0 ], [ 0, 0, 4 ] ] ); Mon( (1,2), [ E(3), 1, 4 ] )
DegreeMon( m )
DegreeMon
returns the degree of the mon m. The degree is the size
of the represented matrix.
gap> DegreeMon( Mon( (1,2), [1, 2, 3] ) ); 3
CharacteristicMon( m )
CharacteristicMon
returns the characteristic of the field
from which the components of the mon m are.
gap> CharacteristicMon( Mon( [1,2] ) ); 0 gap> CharacteristicMon( Mon( (1,2), 4, 5) ); 5
OrderMon( m )
OrderMon
returns the order of the monomial matrix represented by the
mon m. The order of m is the least positive integer r such that
<m>r is the identity. Note that the order might be infinite.
gap> OrderMon( Mon( [1,2] ) ); "infinity" gap> OrderMon( Mon( (1,2), [1, E(3), E(3)^2] ) ); 6
TransposedMon( m )
TransposedMon
returns a mon representing the transposed monomial
matrix of the mon m.
gap> TransposedMon( Mon( [1,2] ) ); Mon( [ 1, 2 ] ) gap> TransposedMon( Mon( (1,2,3), 4 ) ); Mon( (1,3,2), 4 )
DeterminantMon( m )
DeterminantMon
returns the determinant of the monomial matrix
represented by the mon m.
gap> DeterminantMon( Mon( (1,2), [1, E(3), E(3)^2] ) ); -1 gap> DeterminantMon( Mon( [1,2] ) ); 2
TraceMon( m )
TraceMon
returns the trace of the monomial matrix
represented by the mon m.
gap> TraceMon( Mon( (1,2), 4, 5) ); Z(5) gap> TraceMon( Mon( [1,2] ) ); 3
GaloisMon( m, aut )
GaloisMon( m, k )
GaloisMon
returns a mon which is a galois conjugate of the mon m.
This means that each component of the represented matrix is mapped
with an automorphism of the underlying field.
The conjugating automorphism may either be a field automorphism aut
or an integer k specifying the automorphism x -> GaloisCyc(x, k)
in the case characteristic = 0 or x -> x^(FrobeniusAut^k)
in the
case characteristic = p prime.
gap> GaloisMon( Mon( (1,2), [1, E(3), E(3)^2] ), -1 ); Mon( (1,2), [ 1, E(3)^2, E(3) ] ) gap> aut := FrobeniusAutomorphism( GF(4) ); FrobeniusAutomorphism( GF(2^2) ) gap> GaloisMon( Mon( [ Z(2)^0, Z(2^2), Z(2^2)^2 ] ), aut ); Mon( [ Z(2)^0, Z(2^2)^2, Z(2^2) ] )
DirectSumMon( m1, ..., mk )
DirectSumMon
returns the direct sum of the mons
m1, ..., mk. The direct sum of mons is defined via
the direct sum of the represented matrices.
Note that the mons must have the same characteristic.
gap> m1 := Mon( (1,2), [1, E(3), E(3)^2] ); Mon( (1,2), [ 1, E(3), E(3)^2 ] ) gap> m2 := Mon( (1,2), 3); Mon( (1,2), 3 ) gap> DirectSumMon( m1, m2 ); Mon( (1,2)(4,5), [ 1, E(3), E(3)^2, 1, 1, 1 ] )
DirectSumMon( list )
DirectSumMon
returns a mon representing the direct sum of the mons
in list.
gap> m1 := Mon( (1,2), [1, E(3), E(3)^2] ); Mon( (1,2), [ 1, E(3), E(3)^2 ] ) gap> m2 := Mon( (1,2), 3); Mon( (1,2), 3 ) gap> DirectSumMon( [m1, m2] ); Mon( (1,2)(4,5), [ 1, E(3), E(3)^2, 1, 1, 1 ] )
TensorProductMon( m1, ..., mk )
TensorProductMon
returns the tensor product of the mons
m1, ..., mk. The tensor product of mons is defined via
the tensor product (or Kronecker product) of the represented matrices.
Note that the mons must have the same characteristic.
gap> m1 := Mon( (1,2), [1, E(3), E(3)^2] ); Mon( (1,2), [ 1, E(3), E(3)^2 ] ) gap> m2 := Mon( (1,2), 3); Mon( (1,2), 3 ) gap> TensorProductMon( m1, m2 ); Mon( (1,5)(2,4)(3,6)(7,8), [ 1, 1, 1, E(3), E(3), E(3), E(3)^2, E(3)^2, E(3)^2 ] )
TensorProductMon( list )
TensorProductMon
returns a mon representing the tensor product
of the mons in list.
gap> m1 := Mon( (1,2), [1, E(3), E(3)^2] ); Mon( (1,2), [ 1, E(3), E(3)^2 ] ) gap> m2 := Mon( (1,2), 3); Mon( (1,2), 3 ) gap> TensorProductMon( [m1, m2] ); Mon( (1,5)(2,4)(3,6)(7,8), [ 1, 1, 1, E(3), E(3), E(3), E(3)^2, E(3)^2, E(3)^2 ] )
CharPolyCyclesMon( m )
CharPolyCyclesMon
returns the sorted list of the characteristic
polynomials of the cycles of the mon m. All polynomials are written
in a common polynomial ring. Applying Product
to the result yields
the characteristic polynomial of m.
gap> CharPolyCyclesMon( Mon( (1,2), 3 ) ); [ X(Rationals) - 1, X(Rationals)^2 - 1 ] gap> CharPolyCyclesMon( Mon( (1,2), [1, E(3), E(3)^2] ) ); [ X(CF(3)) + (-E(3)^2), X(CF(3))^2 + (-E(3)) ]
The class AMat ( Abstract Matrices)
is created to represent and calculate efficiently with
structured matrices like e.g.
2.(A⊕ B)C⊗ D. E2, where A, B, C, D, E are
matrices of compatible size/characteristic and ⊕, ⊗ denote
the direct sum and tensor product (Kronecker product) resp.
of matrices. The elements of the class AMat are called
``amats" and implement a recursive datastructure to form
expressions like the one above. Basic constructors for amats
allow to create permutation matrices (see AMatPerm
, AMatPerm),
monomial matrices (see AMatMon
, AMatMon) and
general matrices (see AMatMat
, AMatMat) in an efficient way (e.g. a
permutation matrix is defined by a permutation, the degree
and the characteristic). Higher constructors allow to construct
direct sums (see DirectSumAMat
, DirectSumAMat),
tensor products (see TensorProductAMat
, TensorProductAMat)
etc. from given amats.
Note that while building up a highly structured amat from other amats
no computation is done beside checks for compatibility.
To obtain the matrix represented by an amat the appropiate
function has to be applied (e.g. MatAMat
, MatAMat).
Some remarks on the design of AMat: The class AMat is
what is called a term algebra for expressions representing highly
structured matrices over certain base fields. Amats are not
necessarily square but can also be rectangular. Hence, if an amat
must be invertible (e.g. when it shall conjugate another amat) this
has to be proven by computation. To avoid many of these calculations
the result (the inverse) is stored in the object and many functions
accept a ``hint". E.g. by supplying the hint ``invertible" in the
example above the explicit check for invertibility is suppressed.
Using and passing correct hints is essential for efficient computation.
A common problem in the design of non-trivial term algebras is the
simplification strategy: Aggressive or conservative simplification?
Our approach here is extremely conservative. This means even trivial
subexpressions like 1*A are not automatically simplified.
This allows the user to write functions that return their result
always in a fixed structure, e.g. the result is always a conjugated
direct sum of tensor products even though the conjugation might be
trivial.
Finally, note that amats and normal matrices (i.e. lists of lists
of field elements) do not mix -- you have to convert explicitly
with AMatMat
, MatAMat
etc. This greatly simplifies the
amat module.
We define an amat recursively in Backus-Naur-Form as the disjoint union of the following cases.
\raggedbottom
beg-tabularlllll
\multicolumn5lamat ::=
; &\multicolumn4latomic cases
& & perm & ; & ``perm" (invertible)
& | & mon & ; & ``mon" (invertible)
& | & mat & ; & ``mat"
; &\multicolumn4lcomposed cases
& | & scalar . amat & ; & ``scalarMultiple"
& | & amat . ... . amat & ; & ``product"
& | & amat \^ int & ; & ``power"
& | & amat \^ amat & ; & ``conjugate"
& | & amat ⊕ ... ⊕ amat & ; & ``directSum"
& | & amat ⊗ ... ⊗ amat& ; & ``tensorProduct"
& | & GaloisConjugate(amat, aut) & ; & ``galoisConjugate".
end-tabular
An amat A is a record with at least the following fields:
isAMat
& := & true
operations
& := & AMatOps
type
& : & a string identifying the type of Adimensions
& : & size of the matrix represented ( = [rows, columns] )char
& : & characteristic of the base field
\endtabular
The cases as stated above are distinguished by the field .type
of an amat. Depending on the type additional fields are mandatory
as follows:
beg-tabular{p2.5cmp9cm}
\multicolumn2ltype = "perm":
element
& defining permutation
end-tabular
beg-tabular{p2.5cmp9cm}
\multicolumn2ltype = "mon":
element
& defining mon-object (see Mons)
end-tabular
beg-tabular{p2.5cmp9cm}
\multicolumn2ltype = "mat":
element
& defining matrix (list of lists of field elements)
end-tabular
beg-tabular{p2.5cmp9cm}
\multicolumn2ltype = "scalarMultiple":
element
& the AMat multiplied
scalar
& the scalar
end-tabular
beg-tabular{p2.5cmp9cm}
\multicolumn2ltype = "product":
factors
& list of AMats of compatible dimensions
and the same characteristic
end-tabular
beg-tabular{p2.5cmp9cm}
\multicolumn2ltype = "power":
element
& the square AMat to be raised to exponent
exponent
& the exponent (an integer)
end-tabular
beg-tabular{p2.5cmp9cm}
\multicolumn2ltype = "conjugate":
element
& the square AMat to be conjugated
conjugation
& the conjugating invertible AMat
end-tabular
beg-tabular{p2.5cmp9cm}
\multicolumn2ltype = "directSum":
summands
& List of AMats of the same characteristic
end-tabular
beg-tabular{p2.5cmp9cm}
\multicolumn2ltype = "tensorProduct":
factors
& List of AMats of the same characteristic
end-tabular
beg-tabular{p2.5cmp9cm}
\multicolumn2ltype = "galoisConjugate":
element
& the AMat to be Galois conjugated
galoisAut
& the Galois automorphism
end-tabular
Note that there is an important difference between the type of an amat and the type of the matrix being represented by the amat: An amat can be of type ``mat" but the matrix is in fact a permutation matrix. This distinction is refelcted in the naming of the functions: ``XAMat" refers to the type of the amat, ``XMat" to the type of the matrix being represented,
Here a short overview of the functions concerning amats. sections AMatPerm -- Comparison of AMats are concerned with the construction of amats, sections Converting AMats -- MatAMatAMat with the convertability and conversion of amats to permutations, mons and matrices, sections Functions for AMats -- UpperBoundLinearComplexityAMat contain functions for amats, e.g. computation of the determinant or simplification of amats.
The functions concerning amats are implemented in the file
"arep/lib/amat.g"
.
74.23 AMatPerm
AMatPerm( p, d )
AMatPerm( p, d, char )
AMatPerm( p, d, field )
AMatPerm
returns an amat of type "perm"
representing
the (d×d) permutation matrix
[δip j| i,j∈{1,...,d}]
corresponding to the permutation p. As optional parameter
a characteristic char or a field can be supplied.
The default characteristic is zero. The function will signal
an error if the degree d is less than the largest moved point of p.
gap> AMatPerm( (1,2), 5 ); AMatPerm((1,2), 5) gap> AMatPerm( (1,2,3), 5 , 3); AMatPerm((1,2,3), 5, GF(3)) gap> A := AMatPerm( (1,2,3), 5 , Rationals); AMatPerm((1,2,3), 5) gap> A.type; "perm"
AMatMon( m )
AMatMon
returns an amat of type "mon"
representing the
monomial matrix given by the mon m. For the explanation of mons
please refer to Mons.
gap> AMatMon( Mon( (1,2), [1, E(3), E(3)^2] ) ); AMatMon( Mon( (1,2), [ 1, E(3), E(3)^2 ] ) ) gap> A := AMatMon( Mon( (1,2), 3) ); AMatMon( Mon( (1,2), 3 ) ) gap> A.type; "mon"
AMatMat( M )
AMatMat( M, hint )
AMatMat
returns an amat of type "mat"
representing the
matrix M. If the optional hint "invertible"
is supplied
then the field .isInvertible
of the amat is set to true
(without
checking) indicating that the matrix represented is invertible.
gap> AMatMat( [ [1,2], [3,4] ] ); AMatMat( [ [ 1, 2 ], [ 3, 4 ] ] ) gap> A := AMatMat( [ [1,2], [3,4] ] , "invertible"); AMatMat( [ [ 1, 2 ], [ 3, 4 ] ], "invertible" ) gap> A.isInvertible; true
IsAMat( obj )
IsAMat
returns true
if obj, which may be an object of
arbitrary type, is an amat, and false
otherwise.
gap> IsAMat( AMatPerm( (1,2,3), 3 ) ); true gap> IsAMat( 1/2 ); false
IdentityPermAMat( n )
IdentityPermAMat( n, char )
IdentityPermAMat( n, field )
IdentityPermAMat
returns an amat of type "perm"
representing
the (n×n) identity matrix. As optional parameter
a characteristic char or a field can be supplied to obtain
the identity matrix of arbitrary characteristic.
The default characteristic is zero. Note that the same result
can be obtained by using AMatPerm
.
gap> IdentityPermAMat( 3 ); IdentityPermAMat(3) gap> AMatPerm( ( ), 3); IdentityPermAMat(3) gap> IdentityPermAMat( 3 , GF(3) ); IdentityPermAMat(3, GF(3))
IdentityMonAMat( n )
IdentityMonAMat( n, char )
IdentityMonAMat( n, field )
IdentityMonAMat
returns an amat of type "mon"
representing
the (n×n) identity matrix. As optional parameter
a characteristic char or a field can be supplied to obtain
the identity matrix of arbitrary characteristic.
The default characteristic is zero. Note that the same result
can be obtained by using AMatMon
.
gap> IdentityMonAMat( 3 ); IdentityMonAMat(3) gap> AMatMon( Mon( ( ), 3 ) ); IdentityMonAMat(3) gap> IdentityMonAMat( 3, 3 ); IdentityMonAMat(3, GF(3))
IdentityMatAMat( n )
IdentityMatAMat( n, char )
IdentityMatAMat( n, field )
IdentityMatAMat
returns an amat of type "mat"
representing
the (n×n) identity matrix. As optional parameter
a characteristic char or a field can be supplied to obtain
the identity matrix of arbitrary characteristic.
The default characteristic is zero. Note that the same result
can be obtained by using AMatMat
.
gap> IdentityMatAMat( 3 ); IdentityMatAMat(3) gap> AMatMat( [ [1, 0, 0], [0, 1, 0], [0, 0, 1] ]); IdentityMatAMat(3) gap> IdentityMatAMat( 3, GF(3) ); IdentityMatAMat(3, GF(3))
IdentityMatAMat( dim )
IdentityMatAMat( dim, char )
IdentityMatAMat( dim, field )
Let dim be a pair of positive integers. IdentityMatAMat
returns
an amat of type "mat"
representing the rectangular identity
matrix with dim[1]
rows and dim[2]
columns. A rectangular
identity matrix has the entry 1 at the position (i,j) if i = j
and 0 else. As optional parameter a characteristic char or a field
can be supplied to obtain the identity matrix of arbitrary characteristic.
The default characteristic is zero.
gap> IdentityMatAMat( [2, 3] ); IdentityMatAMat([ 2, 3 ]) gap> IdentityMatAMat( [2, 3], 3 ); IdentityMatAMat([ 2, 3 ], GF(3))
IdentityAMat( dim )
IdentityAMat( dim, char )
IdentityAMat( dim, field )
Let dim be a pair of positive integers. IdentityAMat
returns an
amat of type "perm"
if dim[1] = dim[2]
and an amat of type
"mat"
else, representing the identity matrix with dim[1]
rows
and dim[2]
columns. A rectangular identity matrix has the entry 1
at the position (i,j) if i = j and 0 else. Use this function if
you do not know whether the matrix is square and you do not care about
the type. As optional parameter a characteristic char or a field
can be supplied to obtain the identity matrix of arbitrary
characteristic. The default characteristic is zero.
gap> IdentityAMat( [2, 2] ); IdentityPermAMat(2) gap> IdentityAMat( [2, 3] ); IdentityMatAMat([ 2, 3 ])
AllOneAMat( n )
AllOneAMat( n, char )
AllOneAMat( n, field )
AllOneAMat
returns an amat of type "mat"
representing
the (n×n) all-one matrix. An all-one matrix has
the entry 1 at each position. As optional parameter
a characteristic char or a field can be supplied to obtain
the all-one matrix of arbitrary characteristic.
The default characteristic is zero.
gap> AllOneAMat( 3 ); AllOneAMat(3) gap> AllOneAMat( 3, 3); AllOneAMat(3, GF(3))
AllOneAMat( dim )
AllOneAMat( dim, char )
AllOneAMat( dim, field )
Let dim a pair of positive integers. AllOneAMat
returns
an amat of type "mat"
representing the rectangular
all-one matrix with dim[1]
rows and dim[2]
columns.
As optional parameter a characteristic char or a field
can be supplied to obtain the all-one matrix of arbitrary
characteristic. The default characteristic is zero.
gap> AllOneAMat( [3, 2] ); AllOneAMat([ 3, 2 ]) gap> AllOneAMat( [3, 2], GF(5) ); AllOneAMat([ 3, 2 ], GF(5))
NullAMat( n )
NullAMat( n, char )
NullAMat( n, field )
NullAMat
returns an amat of type "mat"
representing
the (n×n) all-zero matrix. An all-zero matrix has
the entry 0 at each position. As optional parameter
a characteristic char or a field can be supplied to obtain
the all-zero matrix of arbitrary characteristic.
The default characteristic is zero.
gap> NullAMat( 3 ); NullAMat(3) gap> NullAMat( 3, 3); NullAMat(3, GF(3))
NullAMat( dim )
NullAMat( dim, char )
NullAMat( dim, field )
Let dim a pair of positive integers. NullAMat
returns
an amat of type "mat"
representing the rectangular
all-zero matrix with dim[1]
rows and dim[2]
columns.
As optional parameter a characteristic char or a field
can be supplied to obtain the all-zero matrix of arbitrary characteristic.
The default characteristic is zero.
gap> NullAMat( [3, 2] ); NullAMat([ 3, 2 ]) gap> NullAMat( [3, 2], GF(5) ); NullAMat([ 3, 2 ], GF(5))
DiagonalAMat( list )
Let list contain field elements of the same characteristic.
DiagonalAMat
returns an amat representing the diagonal matrix
with diagonal entries in list. If all elements in list are
≠ 0 the returned amat is of type "mon"
, else of type
"directSum"
(see AMats).
gap> DiagonalAMat( [2, 3] ); DiagonalAMat([ 2, 3 ]) gap> DiagonalAMat( [0, 2, 3] ); DirectSumAMat( NullAMat(1), AMatMat( [ [ 2 ] ] ), AMatMat( [ [ 3 ] ] ) )
DFTAMat( n )
DFTAMat( n, char )
DFTAMat( n, field )
DFTAMat
returns a special amat of type "mat"
representing the
matrix
DFTn = [ωn i. j| i,j∈{0,...,n-1}], |
Size( F )
.
If these conditions are violated an error is signaled.
The choice of ωn is E(n)
if <char> = 0 and
Z(q)^((q-1)/n)
for <char> = p, q an appropiate p-power.
gap> DFTAMat(3); DFTAMat(3) gap> DFTAMat(3, 7); DFTAMat(3, 7)
SORAMat( n )
SORAMat( n, char )
SORAMat( n, field )
SORAMat
returns a special amat of type "mat"
representing the
matrix
SORn = [ |
| ]. |
gap> SORAMat( 4 ); SORAMat(4) gap> SORAMat( 4, 7); SORAMat(4, 7)
ScalarMultipleAMat( s, A )
or s * A
Let s be a field element and A an amat.
ScalarMultipleAMat
returns an amat of type "scalarMultiple"
representing the scalar multiple of s with A, which must have
common characteristic otherwise an error is signaled.
Note that s and A can be accessed in the fields .scalar
resp. .element
of the result.
gap> A := AMatPerm( (1,2,3), 4); AMatPerm((1,2,3), 4) gap> ScalarMultipleAMat( E(3), A ); E(3) * AMatPerm((1,2,3), 4) gap> 2 * A; 2 * AMatPerm((1,2,3), 4)
74.37 Product and Quotient of AMats
A * B
Let A and B be amats. A * B
returns an amat of
type "product"
representing the product of A and B,
which must have compatible sizes and common characteristic
otherwise an error is signaled.
Note that the factors can be accessed in the field
.factors
of the result.
gap> A := AMatPerm( (1,2,3), 4); AMatPerm((1,2,3), 4) gap> B := AMatMat( [ [1, 2], [3, 4], [5, 6], [7, 8] ] ); AMatMat( [ [ 1, 2 ], [ 3, 4 ], [ 5, 6 ], [ 7, 8 ] ] ) gap> A * A; AMatPerm((1,2,3), 4) * AMatPerm((1,2,3), 4) gap> C := A * B; AMatPerm((1,2,3), 4) * AMatMat( [ [ 1, 2 ], [ 3, 4 ], [ 5, 6 ], [ 7, 8 ] ] ) gap> C.type; "product"
A / B
Let A and B be amats. A / B
returns an amat of
type "product"
representing the quotient of A and B.
The sizes and characteristics of A and B must be compatible,
B must be square and invertible otherwise an error is signaled.
gap> A := AMatPerm( (1,2,3), 4); AMatPerm((1,2,3), 4) gap> B := DiagonalAMat( [1, E(3), 1, 3] ); DiagonalAMat([ 1, E(3), 1, 3 ]) gap> A / B; AMatPerm((1,2,3), 4) * DiagonalAMat([ 1, E(3), 1, 3 ]) ^ -1
PowerAMat( A, n )
or A ^ n
PowerAMat( A, n, hint )
Let A be an amat and n an integer. PowerAMat
returns
an amat of type "power"
representing the power of A with n.
A must be square otherwise an error is signaled. If n is
negative then A is checked for invertibility if the
hint "invertible"
is not supplied.
Note that A and n can be accessed in the fields
.element
resp. .exponent
of the result.
gap> A := AMatPerm( (1,2,3), 4); AMatPerm((1,2,3), 4) gap> B := PowerAMat(A, 3); AMatPerm((1,2,3), 4) ^ 3 gap> B ^ -2; ( AMatPerm((1,2,3), 4) ^ 3 ) ^ -2
ConjugateAMat( A, B )
or A ^ B
ConjugateAMat( A, B, hint )
Let A and B be amats. ConjugateAMat
returns
an amat of type "conjugate"
representing the conjugate of
A with B (i.e. the matrix defined by <B>-1.A.B).
A and B must be square otherwise an error is signaled.
B is checked for invertibility if the
hint "invertible"
is not supplied.
Note that A and B can be accessed in the fields .element
resp. conjugation
of the result.
gap> A := AMatMon( Mon( (1,2), [1, E(4), -1] ) ); AMatMon( Mon( (1,2), [ 1, E(4), -1 ] ) ) gap> B := DFTAMat( 3 ); DFTAMat(3) gap> ConjugateAMat( A, B, "invertible" ); ConjugateAMat( AMatMon( Mon( (1,2), [ 1, E(4), -1 ] ) ), DFTAMat(3) ) gap> B ^ SORAMat( 3 ); ConjugateAMat( DFTAMat(3), SORAMat(3) )
DirectSumAMat( A1, ..., Ak )
DirectSumAMat
returns an amat of type "directSum"
representing
the direct sum of the amats A1, ..., Ak, which must have
common characteristic otherwise an error is signaled. Note that the direct
summands can be accessed in the field .summands
of the result.
gap> A1 := AMatMat( [ [1, 2] ] ); AMatMat( [ [ 1, 2 ] ] ) gap> A2 := DFTAMat( 2 ); DFTAMat(2) gap> A3 := AMatPerm( (1,2), 2 ); AMatPerm((1,2), 2) gap> DirectSumAMat( E(3) * A1, A2 ^ 2, A3 ); DirectSumAMat( E(3) * AMatMat( [ [ 1, 2 ] ] ), DFTAMat(2) ^ 2, AMatPerm((1,2), 2) )
DirectSumAMat( list )
DirectSumAMat
returns an amat of type "directSum"
representing
the direct sum of the amats in list. The amats must have common
characteristic otherwise an error is signaled. The direct
summands can be accessed in the field .summands
of the result.
gap> A := DiagonalAMat( [ Z(3), Z(3)^2 ]); DiagonalAMat([ Z(3), Z(3)^0 ]) gap> B := AMatPerm( (1,2), 3, 3); AMatPerm((1,2), 3, GF(3)) gap> DirectSumAMat( [A, B] ); DirectSumAMat( DiagonalAMat([ Z(3), Z(3)^0 ]), AMatPerm((1,2), 3, GF(3)) )
TensorProductAMat( A1, ..., Ak )
TensorProductAMat
returns an amat of type "tensorProduct"
representing the tensor product (or Kronecker product) of the
amats A1, ..., Ak,
which must have common characteristic otherwise an error is signaled.
Note that the tensor factors can be accessed in the field .factors
of
the result.
gap> A := IdentityPermAMat( 2 ); IdentityPermAMat(2) gap> B := AMatMat( [ [1, 2, 3], [4, 5, 6] ] ); AMatMat( [ [ 1, 2, 3 ], [ 4, 5, 6 ] ] ) gap> TensorProductAMat( A, B ); TensorProductAMat( IdentityPermAMat(2), AMatMat( [ [ 1, 2, 3 ], [ 4, 5, 6 ] ] ) )
TensorProductAMat( list )
TensorPoductAMat
returns an amat of type "tensorProduct"
representing the tensor product of the amats in list.
The amats must have common characteristic otherwise an error
is signaled. The tensor factors can be accessed in the field
.factors
of the result.
gap> A := AMatPerm( (1,2), 3 ); AMatPerm((1,2), 3) gap> B := AMatMat( [ [1], [2] ]); AMatMat( [ [ 1 ], [ 2 ] ] ) gap> TensorProductAMat( [A ^ 2, 2 * B] ); TensorProductAMat( AMatPerm((1,2), 3) ^ 2, 2 * AMatMat( [ [ 1 ], [ 2 ] ] ) )
GaloisConjugateAMat( A, k )
GaloisConjugateAMat( A, aut )
GaloisConjugateAMat
returns an amat which represents a Galois
conjugate of the amat A. The
conjugating automorphism may either be a field automorphism aut
or an integer k specifying the automorphism x -> GaloisCyc(x, k)
in the case characteristic = 0 or x -> x^(FrobeniusAut^k)
in the case characteristic = p prime.
Note that A and k/aut can be accessed in the fields .element
resp.
.galoisAut
of the result.
gap> A := DiagonalAMat( [1, E(3)] ); DiagonalAMat([ 1, E(3) ]) gap> GaloisConjugateAMat( A, -1 ); GaloisConjugateAMat( DiagonalAMat([ 1, E(3) ]), -1 ) gap> aut := FrobeniusAutomorphism( GF(4) ); FrobeniusAutomorphism( GF(2^2) ) gap> B := AMatMon( Mon( (1,2), [ Z(2)^0, Z(2^2) ] ) ); AMatMon( Mon( (1,2), [ Z(2)^0, Z(2^2) ] ) ) gap> GaloisConjugateAMat( B, aut ); GaloisConjugateAMat( AMatMon( Mon( (1,2), [ Z(2)^0, Z(2^2) ] ) ), FrobeniusAutomorphism( GF(2^2) ) )
A = B
A <> B
The equality operator =
evaluates to true
if the amats A
and B are equal and to false
otherwise. The inequality operator
<>
evaluates to true
if the amats A and B are not equal
and to false
otherwise.
Two amats are equal iff they define the same matrix.
gap> A := DiagonalAMat( [E(3), 1] ); DiagonalAMat([ E(3), 1 ]) gap> B := A ^ 3; DiagonalAMat([ E(3), 1 ]) ^ 3 gap> B = IdentityPermAMat( 2 ); true
A < B
A <= B
A >= B
A > B
The operators <
, <=
, >=
, and >
evaluate to true
if the
amat A is strictly less than, less than or equal to, greater than or
equal to, and strictly greater than the amat B.
The ordering of amats is defined via the ordering
of records.
74.44 Converting AMats
The following sections describe the functions for the convertability and conversion of amats to permutations, mons (see Mons) and matrices.
The names of the conversion functions are chosen according to the usual
GAP3-convention: ChalkCheese
makes chalk from cheese.
The parts in the name (chalk, cheese) are
IsIdentityMat( A )
IsIdentityMat
returns true
if the matrix represented by
the amat A is the identity matrix and false
otherwise.
Note that the name of the function
is not IsIdentityAMat
since A can be of any type but represents an
identity matrix in the mathematical sense.
gap> IsIdentityMat(AMatPerm( (1,2), 3 )); false gap> A := DiagonalAMat( [Z(3), Z(3)] ) ^ 2; DiagonalAMat([ Z(3), Z(3) ]) ^ 2 gap> IsIdentityMat(A); true
IsPermMat( A )
IsPermMat
returns true
if the matrix represented by the amat A is
a permutation matrix and false
otherwise. The name of the function
is not IsPermAMat
since A can be of any type but represents a
permutation matrix in the mathematical sense.
Note that IsPermMat
sets and tests A.isPermMat
.
gap> IsPermMat( AMatMon( Mon( (1,2), [1, -1] ))); false gap> IsPermMat( DiagonalAMat( [Z(3), Z(9)] ) ^ 8); true
IsMonMat( A )
IsMonMat
returns true
if the matrix represented by the amat A is
a monomial matrix (a matrix containing exactly one entry ≠ 0
in every row and column) and false
otherwise.
The name of the function is not IsMonAMat
since A can be of any type but represents a monomial matrix
in the mathematical sense. Note that IsMonMat
sets and tests
A.isMonMat
.
gap> IsMonMat( AMatPerm( (1,2), 3 )); true gap> IsMonMat( AMatPerm( (1,2,3), 3 ) ^ DFTAMat(3) ); true
PermAMat( A )
Let A be an amat. PermAMat
returns the permutation represented
by A if A is a permutation matrix (i.e. IsPermMat( A ) = true
)
and false
otherwise. Note that PermAMat
sets and tests A.perm
.
gap> PermAMat(AMatPerm( (1,2), 5 )); (1,2) gap> A := AMatMat( [ [Z(3)^0, Z(3)], [0*Z(3), Z(3)^0] ] ); AMatMat( [ [ Z(3)^0, Z(3) ], [ 0*Z(3), Z(3)^0 ] ] ) gap> PermAMat(A); false gap> PermAMat(A ^ 3); ()
MonAMat( A )
Let A be an amat. MonAMat
returns the mon (see Mons) represented
by A if A is a monomial matrix (i.e. IsMonMat( A ) = true
)
and false
otherwise. Note that MonAMat
sets and tests A.mon
.
gap> MonAMat(AMatPerm( (1,2,3), 5 )); Mon( (1,2,3), 5 ) gap> MonAMat(AMatPerm( (1,2,3), 3 ) ^ DFTAMat(3) ); Mon( [ 1, E(3), E(3)^2 ] ) gap> MonAMat( AMatMat( [ [1, 2] ] )); false
MatAMat( A )
MatAMat
returns the matrix represented by the amat A.
Note that MatAMat
sets and tests A.mat
.
gap> MatAMat( AMatPerm( (1,2), 3, 2 )); [ [ 0*Z(2), Z(2)^0, 0*Z(2) ], [ Z(2)^0, 0*Z(2), 0*Z(2) ], [ 0*Z(2), 0*Z(2), Z(2)^0 ] ] gap> MatAMat(DFTAMat(3)); [ [ 1, 1, 1 ], [ 1, E(3), E(3)^2 ], [ 1, E(3)^2, E(3) ] ] gap> A := IdentityPermAMat(2); IdentityPermAMat(2) gap> B := AMatMat( [ [1,2], [3,4] ] ); AMatMat( [ [ 1, 2 ], [ 3, 4 ] ] ) gap> MatAMat(TensorProductAMat(A, B)); [ [ 1, 2, 0, 0 ], [ 3, 4, 0, 0 ], [ 0, 0, 1, 2 ], [ 0, 0, 3, 4 ] ]
PermAMatAMat( A )
Let A be an amat. PermAMatAMat
returns an amat of type "perm"
equal to A if A is a permutation matrix
(i.e. IsPermMat( A ) = true
) and false
otherwise.
gap> PermAMatAMat(AMatMon(Mon( (1,2), 3 ))); AMatPerm((1,2), 3) gap> PermAMatAMat(DiagonalAMat( [E(3), 1] ) ^ 3); IdentityPermAMat(2) gap> PermAMatAMat(AMatMat( [ [1,2] ] )); false
MonAMatAMat( A )
Let A be an amat. MonAMatAMat
returns an amat of type "mon"
equal to A if A is a monomial matrix (i.e. IsMonMat( A ) = true
)
and false
otherwise.
gap> MonAMat(AMatPerm( (1,2), 3 )); Mon( (1,2), 3 ) gap> MonAMat(DFTAMat(3)^2); Mon( (2,3), [ 3, 3, 3 ] ) gap> MonAMat(AMatMat( [ [1, 2] ] )); false
MatAMatAMat( A )
MatAMatAMat
returns an amat of type "mat"
equal to A.
gap> A := AMatPerm( (1,2), 2 ); AMatPerm((1,2), 2) gap> B := AMatMat( [ [1,2] ] ); AMatMat( [ [ 1, 2 ] ] ) gap> MatAMatAMat(DirectSumAMat(A, B)); AMatMat( [ [ 0, 1, 0, 0 ], [ 1, 0, 0, 0 ], [ 0, 0, 1, 2 ] ] )
The following sections describe useful functions for the calculation with amats (e.g. calculation of the inverse, determinant of an amat as well as simplifying amats). Most of these functions can take great advantage of the highly structured form of the amats.
InverseAMat( A )
InverseAMat
returns an amat representing the inverse of the amat A.
If A is not invertible, an error is signaled. The function uses
mathematical rules to invert the direct sum, tensor product etc. of
matrices. Note that InverseAMat
sets and tests A.inverse
.
gap> A := AMatPerm( (1,2), 3); AMatPerm((1,2), 3) gap> B := AMatMat( [ [1,2], [3,4] ]); AMatMat( [ [ 1, 2 ], [ 3, 4 ] ] ) gap> C := DiagonalAMat( [ E(3), 1] ); DiagonalAMat([ E(3), 1 ]) gap> D := DirectSumAMat(A, TensorProductAMat(B, C)); DirectSumAMat( AMatPerm((1,2), 3), TensorProductAMat( AMatMat( [ [ 1, 2 ], [ 3, 4 ] ] ), DiagonalAMat([ E(3), 1 ]) ) ) gap> InverseAMat(D); DirectSumAMat( AMatPerm((1,2), 3), TensorProductAMat( AMatMat( [ [ -2, 1 ], [ 3/2, -1/2 ] ], "invertible" ), DiagonalAMat([ E(3)^2, 1 ]) ) )
TransposedAMat( A )
TransposedAMat
returns an amat representing the transpose of the
amat A.
The function uses mathematical rules to transpose the direct sum,
tensor product etc. of matrices.
gap> A := AMatPerm( (1,2,3), 3); AMatPerm((1,2,3), 3) gap> B := AMatMat( [ [1, 2] ] ); AMatMat( [ [ 1, 2 ] ] ) gap> TransposedAMat(TensorProductAMat(A, B)); TensorProductAMat( AMatPerm((1,3,2), 3), AMatMat( [ [ 1 ], [ 2 ] ] ) )
DeterminantAMat( A )
DeterminantAMat
returns the determinant of the amat A. If
A is not square an error is signaled.
The function uses mathematical rules to calculate the determinant of
the direct sum, tensor product etc. of matrices.
Note that DeterminantAMat
sets and tests A.determinant
.
gap> A := AMatMat( [ [1,2], [3,4] ] ); AMatMat( [ [ 1, 2 ], [ 3, 4 ] ] ) gap> B := AMatPerm( (1,2), 2 ); AMatPerm((1,2), 2) gap> DeterminantAMat(TensorProductAMat(A, B)); 4
TraceAMat( A )
TraceAMat
returns the trace of the amat A. If
A is not square an error is signaled.
The function uses mathematical rules to calculate the trace of
direct sums, tensor product etc. of matrices.
Note that TraceAMat
sets and tests A.trace
.
gap> A := DFTAMat(2); DFTAMat(2) gap> B := DiagonalAMat( [1, 2, 3] ); DiagonalAMat([ 1, 2, 3 ]) gap> TraceAMat(DirectSumAMat( A^2, B )); 10
RankAMat( A )
RankAMat
returns the rank of the amat A. Note that RankAMat
sets and tests A.rank
.
gap> RankAMat(AllOneAMat(100)); 1 gap> RankAMat(AMatPerm( (1,2), 10 )); 10
SimplifyAMat( A )
SimplifyAMat
returns a simplified amat representing the same matrix
as the amat A. The simplification is performed recursively according
to certain rules.
E.g. the following simplifications are performed:
• If A represents a permutation matrix, monomial matrix then an amat of type ``perm", ``mon" resp. is returned.
• In a product resp. tensor product, trivial factors are omitted.
• Trivial conjugation is omitted.
• In a direct sum adjacent permutation/monomial matrices are put together.
• In a product adjacent permutation/monomial matrices are multiplied together.
• Successive scalars are multiplied together.
• Successive exponents are multiplied together, negative
exponents are evaluated using InverseAMat
.
Note that important information about the matrix is
shifted to the simplification.
gap> A := IdentityPermAMat( 3 ); IdentityPermAMat(3) gap> B := DiagonalAMat( [E(3), 1, 1] ); DiagonalAMat([ E(3), 1, 1 ]) gap> C := AMatMat( [ [1,2], [3,4] ] ); AMatMat( [ [ 1, 2 ], [ 3, 4 ] ] ) gap> D := DirectSumAMat(A ^ -1, 1 * B * A, C); DirectSumAMat( IdentityPermAMat(3) ^ -1, ( 1 * DiagonalAMat([ E(3), 1, 1 ]) ) * IdentityPermAMat(3), AMatMat( [ [ 1, 2 ], [ 3, 4 ] ] ) ) gap> SimplifyAMat(D); DirectSumAMat( IdentityPermAMat(3), DiagonalAMat([ E(3), 1, 1 ]), AMatMat( [ [ 1, 2 ], [ 3, 4 ] ] ) )
kbsAMat( A1, ..., Ak )
kbsAMat
returns the joined kbs (conjugated blockstructure)
of the amats A1, ..., Ak.
The amats must be square and of common size and characteristic
otherwise an error is signaled. The joined kbs of a list of
(n× n)-matrices is a partition of {1,...,n} representing their
common blockstructure. For an exact definition see kbs.
gap> A := IdentityPermAMat(2); IdentityPermAMat(2) gap> B := AMatMat( [ [1,2], [3,4] ] ); AMatMat( [ [ 1, 2 ], [ 3, 4 ] ] ) gap> kbsAMat(TensorProductAMat(A, B)); [ [ 1, 2 ], [ 3, 4 ] ] gap> kbsAMat(AMatPerm( (1,3)(2,4), 5 )); [ [ 1, 3 ], [ 2, 4 ], [ 5 ] ]
kbsAMat( list )
kbsAMat
returns the joined kbs of the amats in list (see above).
kbsDecompositionAMat( A )
kbsDecompositionAMat
decomposes the amat A into a conjugated
(by an amat of type "perm"
) direct sum of amats of type "mat"
as far as possible. If A is not square an error is signaled.
The decomposition is performed according to the
kbs (see kbs) of A which is a partition of {1,...,n}
(n = number of rows of A) describing the blockstructure of A.
gap> A := AMatMat( [[1,0,2,0], [0,1,0,2], [3,0,4,0], [0,3,0,4]] ); AMatMat( [ [ 1, 0, 2, 0 ], [ 0, 1, 0, 2 ], [ 3, 0, 4, 0 ], [ 0, 3, 0, 4 ] ] ) gap> kbsDecompositionAMat(A); ConjugateAMat( DirectSumAMat( AMatMat( [ [ 1, 2 ], [ 3, 4 ] ] ), AMatMat( [ [ 1, 2 ], [ 3, 4 ] ] ) ), AMatPerm((2,3), 4) )
AMatSparseMat( M )
AMatSparseMat( M, match-blocks )
Let M be a sparse matrix (i.e. containing entries ≠ 0).
AMatSparseMat
returns an amat of the form <P>1. E1.
D. E2. P2 where (for i = 1,2) <P>i are amats of type
"perm"
, <E>i are identity-amats (might be rectangular) and D
is an amat of type "directSum"
. If match-blocks is true
or
not provided then, furthermore, the permutations p1 and p2
are chosen such that equivalent summands of D are equal and
collected together by a tensor product.
If match-blocks is false
this is not done.
The major part of the work is done by the function
DirectSummandsPermutedMat
(see DirectSummandsPermutedMat).
Use the function SimplifyAMat
(see SimplifyAMat) for
simplification of the result.
For an explanation of the algorithm see Egn97.
gap> M := [[0,0,0,0],[0,1,0,2],[0,0,3,0],[0,4,0,5]];; gap> PrintArray(M); [ [ 0, 0, 0, 0 ], [ 0, 1, 0, 2 ], [ 0, 0, 3, 0 ], [ 0, 4, 0, 5 ] ] gap> AMatSparseMat(M); AMatPerm((1,4,3), 4) * IdentityMatAMat([ 4, 3 ]) * DirectSumAMat( TensorProductAMat( IdentityPermAMat(1), AMatMat( [ [ 3 ] ] ) ), TensorProductAMat( IdentityPermAMat(1), AMatMat( [ [ 1, 2 ], [ 4, 5 ] ] ) ) ) * IdentityMatAMat([ 3, 4 ]) * AMatPerm((1,3,4), 4)
SubmatrixAMat( A, inds )
Let A be an amat and inds a set of positive integers.
SubmatrixAMat
returns an amat of type "mat"
representing
the submatrix of A defined by extracting all entries with
row and column index in inds.
gap> A := AMatPerm( (1,2), 2 ); AMatPerm((1,2), 2) gap> B := AMatMat( [ [1,2], [3,4] ] ); AMatMat( [ [ 1, 2 ], [ 3, 4 ] ] ) gap> SubmatrixAMat(TensorProductAMat(A, B), [2,3] ); AMatMat( [ [ 0, 3 ], [ 2, 0 ] ] )
74.65 UpperBoundLinearComplexityAMat
UpperBoundLinearComplexityAMat( A )
UpperBoundLinearComplexityAMat
returns an upper bound for the linear
complexity of the amat A according to the complexity model
L∞ of Clausen/Baum, CB93. The linear complexity is
a measure for the complexity of the matrix-vector multiplication of a
given matrix with an arbitrary vector.
gap> UpperBoundLinearComplexityAMat(DFTAMat(2)); 2 gap> UpperBoundLinearComplexityAMat(DiagonalAMat( [2, 3] )); 2 gap> A := AMatPerm( (1,2), 3); AMatPerm((1,2), 3) gap> B := AMatMat( [ [1,2], [3,4] ] ); AMatMat( [ [ 1, 2 ], [ 3, 4 ] ] ) gap> UpperBoundLinearComplexityAMat(TensorProductAMat(A, B)); 24
The class ARep ( Abstract Representations) is created to represent and calculate efficiently with structured matrix representations of finite groups up to equality, e.g. expressions like (φ\uparrowT G)M⊗ψ where φ, ψ are representations and \uparrow, ⊗ denotes the induction resp. inner tensor product of representations. The implementation idea is the same as with the class AMat (see AMats), i.e. a representation is a record containing the necessary information (e.g. degree, characteristic, list of images on the generators) to define a representation up to equality. The elements of ARep are called ``areps" and are no group homomorphisms in the sense of GAP3 (which is the reason for the term ``abstract" representation). Special care is taken of permutation and monomial representations, which can be represented very efficiently by storing a list of permutations or mons (see Mons) instead of matrices as images on the generators.
Areps can represent representations of any finite group and any characteristic including modular (characteristic divides group size) representations, but most of the higher functions will only work in the non-modular case or even only in the case of characteristic zero. These restrictions are always indicated in the description of the respective function.
Basic constructors allow to create areps, e.g. by supplying
the list of images on the generators (see ARepByImages
,
ARepByImages). Since GAP3 allows the manipulation of
the generators given to construct a group, it is important
for consistency to have a field with generators one can rely on.
This is realized in the function GroupWithGenerators
,
GroupWithGenerators.
Higher constructors allow to construct inductions (see InductionARep
,
InductionARep), direct sums (see DirectSumARep
, DirectSumARep),
inner tensor products (see InnerTensorProductARep
,
InnerTensorProductARep) etc. from given areps.
Some remarks on the design of ARep:
The class ARep is a term algebra for matrix representations
of finite groups (see also AMat, AMats).
The simplification strategy is extremely conservative,
which means that even trivial expressions like
GaloisConjugate(R, id)
are only simplified upon
explicit request.
As in AMat we use the ``hint"-concept extensively
to suppress unnecessary expensive computations of little interest.
The class AMat is used in ARep in three ways:
1. for images under areps, 2. for conjugating matrices
(change of base of the underlying vector space) and 3. for elements
of the intertwining space of two areps.
Note that 3. requires non-invertible or even rectangular matrices
to be represented.
A special point that deserves mentioning is the way in which areps
act as homomorphisms anf how they are defined. Areps are no
GAP3-homomorphisms. We simply did not manage to implement
ARep as a term algebra and as GAP3-homomorphisms
in a relyable and efficient way which avoids maximal confusion.
In addition, working with ARep usually involves many
representations of the same group. This is supported in the
most obvious way by fixing the list of generators used to
create the group (see GroupWithGenerators) and only varying
the list of images. Although this strategy differs from the
approach in GAP3 (which deliberately manipulates the
generating list used to construct the group) it turned out
to be very useful and efficient in the situation at hand.
We define an arep recursively in Backus-Naur-Form as the disjoint union of the following cases.
beg-tabularlllll
\multicolumn5larep ::=
; &\multicolumn4latomic cases
& & perm & ; & ``perm"
& | & mon & ; & ``mon"
& | & mat & ; & ``mat"
; &\multicolumn4lcomposed cases
& | & arep \^ arep & ; & ``conjugate"
& | & arep ⊕ ... ⊕ arep & ; & ``directSum"
& | & arep ⊗ ... ⊗ arep& ; & ``innerTensorProduct"
& | & arep # ... # arep & ; & ``outerTensorProduct"
& | & arep \downarrow subgrp & ; & ``restriction"
& | & arep \uparrow supgrp, transversal & ; & ``induction"
& | & Extension(arep, ext-character) & ; & ``extension"
& | & GaloisConjugate(arep, aut) & ; & ``galoisConjugate"
end-tabular
An arep R is a record with the following fields mandatory to all types of areps.
isARep
& := & true
operations
& := & AMatOps
char
& : & characteristic of the base fielddegree
& : & degree of the representationsource
& : & the group being represented, which must contain.theGenerators
, see GroupWithGeneratorstype
& : & a string identifying the type of R
\endtabular
The cases as stated above are distinguished by the field .type
of an arep R. Depending on the type additional fields are mandatory
as follows.
beg-tabular{p2.5cmp10cm}
\multicolumn2ltype = "perm":
theImages
& list of permutations for the images
of source.theGenerators
end-tabular
beg-tabular{p2.5cmp10cm}
\multicolumn2ltype = "mon":
theImages
& list of mons (see Mons) for the images
of source.theGenerators
end-tabular
beg-tabular{p2.5cmp10cm}
\multicolumn2ltype = "mat":
theImages
& list of matrices for the images
of source.theGenerators
end-tabular
beg-tabular{p2.5cmp10cm}
\multicolumn2ltype = "mat":
rep
& an arep to be conjugated
conjugation
& an amat (see AMats) conjugating rep
end-tabular
beg-tabular{p2.5cmp10cm}
\multicolumn2ltype = "directSum":
summands
& list of areps of the same source and characteristic
end-tabular
beg-tabular{p2.5cmp10cm}
\multicolumn2ltype = "innerTensorProduct":
factors
& list of areps of the same characteristic
end-tabular
beg-tabular{p2.5cmp10cm}
\multicolumn2ltype = "outerTensorProduct":
factors
& list of areps of the same characteristic
end-tabular
beg-tabular{p2.5cmp10cm}
\multicolumn2ltype = "restriction":
rep
& an arep of a supergroup of source
, the group source
& and rep.source
have the same parent group
end-tabular
beg-tabular{p2.5cmp10cm}
\multicolumn2ltype = "induction":
rep
& an arep of a subgroup of source
, the group source
& and rep.source
have the same parent group
transversal
& a right transversal of Cosets(source, rep.source)
end-tabular
beg-tabular{p2.5cmp10cm}
\multicolumn2ltype = "galoisConjugate":
rep
& an arep to be conjugated
galoisAut
& the Galois automorphism
end-tabular
Note that most of the function concerning areps require calculation in the source group. Hence it is most useful to choose aggroups or permutation groups as sources if possible. Furthermore there is an important difference between the type of an arep and the type of the representation being represented by the arep: E.g. an arep can be of type ``induction" but the representation is in fact a permutation representation. This distinction is reflected in the naming of the functions: ``XARep" refers to the type of the arep, ``XRep" to the type of the representation being represented,
Here a short overview of the function concerning areps.
sections GroupWithGenerators -- GaloisConjugateARep are
concerned with the construction of areps, sections
Basic Functions for AReps -- ARepWithCharacter are
concerned with the evaluation of an arep at a point, tests for
equivalence and irreducibility, construction of an arep
with given character etc., sections Converting AReps --
MatARepARep deal with the conversion of areps to areps
of type "perm", "mon", "mat"
.
Sections Higher Functions for AReps -- DecompositionMonRep
provide function for the computation of the intertwining space
of areps and a plenty of functions for monomial areps. The
most important function here is DecompositionMonRep
(see DecompositionMonRep) performing the decomposition of
a monomial arep including the computation of a highly
structured decomposition matrix.
The basic functions concerning areps are implemented in the file
"arep/lib/arep.g"
, the higher functions in
"arep/lib/arepfcts.g"
.
For details on constructive representation theory and the theoretical background of the higher functions please refer to Pue98.
GroupWithGenerators( G )
Let G be a group. GroupWithGenerators
returns G with the
field G.theGenerators
being set to a fixed non-empty generating
set of G. This function is created because GAP3 has the freedom
to manipulate the generators given to construct a group.
Based on the list G.theGenerators
areps can be
constructed, e.g. by the images on that list
(ARepByImages
, ARepByImages). If an arep for a group G
is constructed with the field G.theGenerators
unbound a
warning is signaled and the field is set.
gap> G := Group( (1,2) ); Group( (1,2) ) gap> GroupWithGenerators(G); Group( (1,2) ) gap> G.theGenerators; [ (1,2) ] gap> G := Group( () ); Group( () ) gap> GroupWithGenerators(G); Group( () ) gap> G.theGenerators; [ () ] gap> G.generators; [ ]
GroupWithGenerators( list )
GroupWithGenerators
returns the group G generated by the elements
in list. The field G.theGenerators
is set to list.
For the reason of this function see above.
gap> G := GroupWithGenerators( [ (), (1,2), (1,2,3) ] ); Group( (1,2), (1,2,3) ) gap> G.theGenerators; [ (), (1,2), (1,2,3) ] gap> G.generators; [ (1,2), (1,2,3) ]
TrivialPermARep( G )
TrivialPermARep( G, d )
TrivialPermARep( G, d, char )
TrivialPermARep( G, d, field )
TrivialPermARep
returns an arep of type "perm"
representing
the one representation of the group G of degree d. The default
degree is 1.
As optional parameter a characteristic char or a field
can be supplied to obtain the one representation of arbitrary
characteristic. The default characteristic is zero.
gap> G := GroupWithGenerators( [(1,2), (3,4)] ); Group( (1,2), (3,4) ) gap> TrivialPermARep(G, 2, 3); TrivialPermARep( GroupWithGenerators( [ (1,2), (3,4) ] ), 2, GF(3) ) gap> G := GroupWithGenerators( [(1,2), (3,4)] ); Group( (1,2), (3,4) ) gap> R := TrivialPermARep(G, 2, 3); TrivialPermARep( GroupWithGenerators( [ (1,2), (3,4) ] ), 2, GF(3) ) gap> R.degree; 2 gap> R.char; 3
TrivialMonARep( G )
TrivialMonARep( G, d )
TrivialMonARep( G, d, char )
TrivialMonARep( G, d, field )
TrivialMonARep
returns an arep of type "mon"
representing
the one representation of the group G of degree d. The default
degree is 1.
As optional parameter a characteristic char or a field
can be supplied to obtain the one representation of arbitrary
characteristic. The default characteristic is zero.
gap> G := GroupWithGenerators( [(1,2), (3,4)] ); Group( (1,2), (3,4) ) gap> R := TrivialMonARep(G, 2); TrivialMonARep( GroupWithGenerators( [ (1,2), (3,4) ] ), 2 ) gap> R.theImages; [ Mon( (), 2 ), Mon( (), 2 ) ]
TrivialMatARep( G )
TrivialMatARep( G, d )
TrivialMatARep( G, d, char )
TrivialMatARep( G, d, field )
TrivialMatARep
returns an arep of type "mat"
representing
the one representation of the group G of degree d. The default
degree is 1.
As optional parameter a characteristic char or a field
can be supplied to obtain the one representation of arbitrary
characteristic. The default characteristic is zero.
gap> G := GroupWithGenerators( [(1,2), (3,4)] ); Group( (1,2), (3,4) ) gap> R := TrivialMatARep(G); TrivialMatARep( GroupWithGenerators( [ (1,2), (3,4) ] ) ) gap> R.theImages; [ [ [ 1 ] ], [ [ 1 ] ] ]
RegularARep( G )
RegularARep( G, char )
RegularARep( G, field )
RegularARep
returns an arep of type "induction"
representing
the regular representation of G. The regular representation is
defined (up to equality) by the induction
R = (1{\mathsfE}\uparrowT G) of the trivial representation
(of degree one) of the trivial subgroup \mathsfE of G with
the transversal T being the ordered list of elements of G.
As optional parameter a characteristic char or a field
can be supplied to obtain the regular representation of arbitrary
characteristic. The default characteristic is zero.
gap> G := GroupWithGenerators(SymmetricGroup(3)); Group( (1,3), (2,3) ) gap> RegularARep(G); RegularARep( GroupWithGenerators( [ (1,3), (2,3) ] ) ) gap> RegularARep(G, GF(2)); RegularARep( GroupWithGenerators( [ (1,3), (2,3) ] ), GF(2) )
NaturalARep( G )
NaturalARep( G, d )
NaturalARep( G, d, char )
NaturalARep( G, d, field )
Let G be a mongroup or a matrix group (for mons see Mons).
NaturalARep
returns an arep of type "mon"
or "mat"
resp.
representing the representation given by G,
which means that G is taken as a representation of itself.
For a permutation group G the desired degree d of the
representation has to be supplied. The returned arep is of
type "perm"
. If d is smaller than the
largest moved point of G an error is signaled.
As optional parameter a characteristic char or a field
can be supplied (if G is a permutation group).
Note that a mongroup or a matrix group as source of an arep
slows down most of the calculations with it.
gap> G := GroupWithGenerators( [ (1,2), (1,2,3) ] ); Group( (1,2), (1,2,3) ) gap> R := NaturalARep(G, 4); NaturalARep( GroupWithGenerators( [ (1,2), (1,2,3) ] ), 4 ) gap> R.theImages; [ (1,2), (1,2,3) ] gap> R.degree; 4 gap> G := GroupWithGenerators( [ Mon( (1,2), [E(4), 1] ) ] ); Group( Mon( (1,2), [ E(4), 1 ] ) ) gap> NaturalARep(G); NaturalARep( GroupWithGenerators( [ Mon( (1,2), [ E(4), 1 ] ) ] ) )
ARepByImages( G, list )
ARepByImages( G, list, hint )
ARepByImages( G, list, d )
ARepByImages( G, list, d, hint )
ARepByImages( G, list, d, char )
ARepByImages( G, list, d, field )
ARepByImages( G, list, d, char, hint )
ARepByImages( G, list, d, field, hint )
ARepByImages
allows to construct an arep of the group G
by supplying the list of images on the list G.theGenerators
.
Let list contain mons (see Mons). ARepByImages
returns an arep
of type "mon"
defined by mapping G.theGenerators
elementwise onto list.
Let list contain matrices. ARepByImages
returns an arep
of type "mat"
defined by mapping G.theGenerators
elementwise onto list.
Let list contain permutations. ARepByImages
returns an arep
of type "perm"
and degree d defined by mapping G.theGenerators
elementwise onto list. If d is smaller than the
largest moved point of G an error is signaled.
As optional parameter a characteristic
char or a field can be supplied to obtain an arep of arbitrary
characteristic.
In all cases the hint "hom"
or "faithful"
can be supplied to
indicate that the list of images does define a homomorphism or
even a faithful homomorphism respectively.
If no hint is supplied it is checked
whether the list of images defines a homomorphism.
gap> G := GroupWithGenerators( [(1,2), (1,2,3)] ); Group( (1,2), (1,2,3) ) gap> ARepByImages(G, [ Mon( [-1] ), Mon( [1] ) ] ); ARepByImages( GroupWithGenerators( [ (1,2), (1,2,3) ] ), [ Mon( [ -1 ] ), Mon( (), 1 ) ], "hom" ) gap> L := [ [ [Z(2), Z(2)], [0*Z(2), Z(2)] ], IdentityMat(2, GF(2)) ]; [ [ [ Z(2)^0, Z(2)^0 ], [ 0*Z(2), Z(2)^0 ] ], [ [ Z(2)^0, 0*Z(2) ], [ 0*Z(2), Z(2)^0 ] ] ] gap> ARepByImages(G, L); ARepByImages( GroupWithGenerators( [ (1,2), (1,2,3) ] ), [ [ [ Z(2)^0, Z(2)^0 ], [ 0*Z(2), Z(2)^0 ] ], [ [ Z(2)^0, 0*Z(2) ], [ 0*Z(2), Z(2)^0 ] ] ], GF(2), "hom" ) gap> ARepByImages(G, [ (1,2), () ], 3); ARepByImages( GroupWithGenerators( [ (1,2), (1,2,3) ] ), [ (1,2), () ], 3, # degree "hom" ) gap> ARepByImages(G, [ (1,2), () ], 3, "hom"); ARepByImages( GroupWithGenerators( [ (1,2), (1,2,3) ] ), [ (1,2), () ], 3, # degree "hom" )
ARepByHom( hom )
ARepByHom( hom, d )
ARepByHom( hom, d, char )
ARepByHom( hom, d, char )
Let hom be a homomorphism of a group into a mongroup.
ARepByHom
returns an arep of type "mon"
corresponding
to hom.
Let hom be a homomorphism of a group into a matrix group.
ARepByHom
returns an arep of type "mat"
corresponding
to hom.
Let hom be a homomorphism of a group into a permutation group
and d a positive integer.
ARepByHom
returns an arep of type "perm"
and degree d
corresponding to hom. If d is smaller than the largest moved
point of hom.range
an error is signaled. As optional parameter
a characteristic char or a field can be supplied to obtain an
arep of arbitrary characteristic.
gap> G := GroupWithGenerators(SymmetricGroup(4)); Group( (1,4), (2,4), (3,4) ) gap> phi := IdentityMapping(G); IdentityMapping( Group( (1,4), (2,4), (3,4) ) ) gap> ARepByHom(phi, 4); NaturalARep( GroupWithGenerators( [ (1,4), (2,4), (3,4) ] ), 4 ) gap> H := GroupWithGenerators( [ Mon( [-1] ) ] ); Group( Mon( [ -1 ] ) ) gap> psi := > GroupHomomorphismByImages(G, H, G.generators, [H.1, H.1, H.1]); GroupHomomorphismByImages( Group( (1,4), (2,4), (3,4) ), Group( Mon( [ -1 ] ) ), [ (1,4), (2,4), (3,4) ], [ Mon( [ -1 ] ), Mon( [ -1 ] ), Mon( [ -1 ] ) ] ) gap> ARepByHom(psi); ARepByImages( GroupWithGenerators( [ (1,4), (2,4), (3,4) ] ), [ Mon( [ -1 ] ), Mon( [ -1 ] ), Mon( [ -1 ] ) ], "hom" )
ARepByCharacter( chi )
Let chi be a onedimensional character of a group.
ARepByCharacter
returns a onedimensional arep of
type "mon"
given by chi.
gap> G := GroupWithGenerators( [ (1,2) ] ); Group( (1,2) ) gap> L := Irr(G); [ Character( Group( (1,2) ), [ 1, 1 ] ), Character( Group( (1,2) ), [ 1, -1 ] ) ] gap> ARepByCharacter( L[2] ); ARepByImages( GroupWithGenerators( [ (1,2) ] ), [ Mon( [ -1 ] ) ], "hom" )
ConjugateARep( R, A )
or R ^ A
ConjugateARep( R, A, hint )
Let R be an arep and A an amat (see AMats).
ConjugateARep
returns an arep of type "conjugate"
representing the conjugated representation
<R>A: x→ A-1. R(x). A. The amat
is tested for invertibility if the optional hint
"invertible"
is not supplied. R and A must be compatible
in size and characteristic otherwise an error is signaled.
Note that R and A can be accessed in the fields .rep
and .conjugation
of the result.
gap> G := GroupWithGenerators(SymmetricGroup(4)); Group( (1,4), (2,4), (3,4) ) gap> R := NaturalARep(G, 4); NaturalARep( GroupWithGenerators( [ (1,4), (2,4), (3,4) ] ), 4 ) gap> A := AMatPerm( (1,2,3,4), 4 ); AMatPerm((1,2,3,4), 4) gap> R ^ A; ConjugateARep( NaturalARep( GroupWithGenerators( [ (1,4), (2,4), (3,4) ] ), 4 ), AMatPerm((1,2,3,4), 4) )
DirectSumARep( R1, ..., Rk )
DirectSumARep
returns an arep of type "directSum"
representing the direct sum <R1>⊕...⊕ R_k
of the areps R1, ..., Rk,
which must have common source and characteristic otherwise an error is
signaled.
The direct sum <R> = R_1⊕...⊕ R_k of representations is defined as x→ R_1(x)⊕...⊕ R_k(x).
Note that the summands R1, ..., Rk can be accessed
in the field .summands
of the result.
gap> G := GroupWithGenerators( [(1,2,3,4), (1,3)] ); Group( (1,2,3,4), (1,3) ) gap> R1 := RegularARep(G); RegularARep( GroupWithGenerators( [ (1,2,3,4), (1,3) ] ) ) gap> R2 := ARepByImages(G, [ [[1]], [[-1]] ]); ARepByImages( GroupWithGenerators( [ (1,2,3,4), (1,3) ] ), [ [ [ 1 ] ], [ [ -1 ] ] ], "hom" ) gap> DirectSumARep(R1, R2); DirectSumARep( RegularARep( GroupWithGenerators( [ (1,2,3,4), (1,3) ] ) ), ARepByImages( GroupWithGenerators( [ (1,2,3,4), (1,3) ] ), [ [ [ 1 ] ], [ [ -1 ] ] ], "hom" ) )
DirectSumARep( list )
DirectSumARep
returns an arep of type "directSum"
representing
the direct sum of the areps in list (see above).
74.78 InnerTensorProductARep
InnerTensorProductARep( R1, ..., Rk )
InnerTensorProductARep
returns an arep of type
"innerTensorProduct"
representing the inner tensor product
<R> = R_1⊗...⊗ R_k of the areps R1, ..., Rk,
which must have common source and characteristic otherwise an error is
signaled.
The inner tensor product <R> = R_1⊗...⊗ R_k of representations is defined as x→ R_1(x)⊗...⊗ R_k(x). Note that the inner tensor product yields a representation of the same source (in contrast to the outer tensor product, see OuterTensorProductARep).
Note that the tensor factors R1, ..., Rk can be accessed
in the field .factors
of the result.
gap> G := GroupWithGenerators( [ (1,2), (3,4) ] ); Group( (1,2), (3,4) ) gap> R1 := ARepByImages(G, [ Mon( (1,2), 2 ), Mon( [-1, -1] ) ] ); ARepByImages( GroupWithGenerators( [ (1,2), (3,4) ] ), [ Mon( (1,2), 2 ), Mon( [ -1, -1 ] ) ], "hom" ) gap> R2 := NaturalARep(G, 5); NaturalARep( GroupWithGenerators( [ (1,2), (3,4) ] ), 5 ) gap> InnerTensorProductARep(R1, R2); InnerTensorProductARep( ARepByImages( GroupWithGenerators( [ (1,2), (3,4) ] ), [ Mon( (1,2), 2 ), Mon( [ -1, -1 ] ) ], "hom" ), NaturalARep( GroupWithGenerators( [ (1,2), (3,4) ] ), 5 ) )
InnerTensorProductARep( list )
InnerTensorProductARep
returns an arep of type
"innerTensorProduct"
representing
the inner tensor product of the areps in list (see above).
74.79 OuterTensorProductARep
OuterTensorProductARep( R1, ..., Rk )
OuterTensorProductARep( G, R1, ..., Rk )
OuterTensorProductARep
returns an arep of type
"outerTensorProduct"
representing the outer tensor product
<R> = R_1# ...# R_k of the areps R1, ..., Rk,
which must have common characteristic otherwise an error is
signaled.
The outer tensor product <R> = R_1# ...# R_k of representations is defined as x→ R_1(x)⊗...⊗ R_k(x). Note that the outer tensor product of representations is a representation of the direct product of the sources (in contrast to the inner tensor product, see InnerTensorProductARep).
Using the first version OuterTensorProductARep
returns an arep R with
R.source = DirectProduct(R1.source, ..., Rk.source)
using the GAP3 function DirectProduct
. In the second version
the returned arep has as source the group G which must be the inner
direct product <G> = R_1.source
×...×R_k.source
.
This property is not checked.
Note that the tensor factors R1, ..., Rk can be accessed
in the field .factors
of the result.
gap> G1 := GroupWithGenerators(DihedralGroup(8)); Group( (1,2,3,4), (2,4) ) gap> G2 := GroupWithGenerators( [ (1,2) ] ); Group( (1,2) ) gap> R1 := NaturalARep(G1, 4); NaturalARep( GroupWithGenerators( [ (1,2,3,4), (2,4) ] ), 4 ) gap> R2 := ARepByImages(G2, [ [[-1]] ]); ARepByImages( GroupWithGenerators( [ (1,2) ] ), [ [ [ -1 ] ] ], "hom" ) gap> OuterTensorProductARep(R1, R2); OuterTensorProductARep( NaturalARep( GroupWithGenerators( [ (1,2,3,4), (2,4) ] ), 4 ), ARepByImages( GroupWithGenerators( [ (1,2) ] ), [ [ [ -1 ] ] ], "hom" ) )
RestrictionARep( R, H )
RestrictionARep
returns an arep of type "restriction"
representing the restriction of the arep R to
the subgroup H of R.source
. Here, ``subgroup" means,
that all elements of H are contained in R.source
.
The restriction <R>\downarrow H of a representation R to a subgroup H is defined by x→ R(x), x∈ H.
Note that R can be accessed in the field .rep
of the result.
gap> G := GroupWithGenerators(SymmetricGroup(4)); Group( (1,4), (2,4), (3,4) ) gap> H := GroupWithGenerators(AlternatingGroup(4)); Group( (1,2,4), (2,3,4) ) gap> R := NaturalARep(G, 4); NaturalARep( GroupWithGenerators( [ (1,4), (2,4), (3,4) ] ), 4 ) gap> RestrictionARep(R, H); RestrictionARep( NaturalARep( GroupWithGenerators( [ (1,4), (2,4), (3,4) ] ), 4 ), GroupWithGenerators( [ (1,2,4), (2,3,4) ] ) )
InductionARep( R, G )
InductionARep( R, G, T )
InductionARep
returns an arep of type "induction"
representing the induction of the arep R to
the supergroup G with the transversal T of the residue classes
R.source
\ G. Here, ``supergroup" means
that all elements of R.source
are contained in G. If no
transversal T is supplied one is chosen by the function
RightTransversal
. If a transversal T is given it is not checked
to be one.
The induction <R>\uparrowT G of a representation R of H to a supergroup G with transversal T = {t1,...,tk} of <H> \ G is defined by x→[ \dotR(ti. x. tj-1) | i,j∈{1,...,k}], where \dotR(y) = R(y) for y∈ H and 0 else.
Note that R and T can be accessed in the fields
.rep
and .transversal
resp. of the result.
gap> G := GroupWithGenerators( [ (1,2,3,4), (1,2) ] ); Group( (1,2,3,4), (1,2) ) gap> H := GroupWithGenerators( [ (1,2) ] ); Group( (1,2) ) gap> R := ARepByImages(H, [ [[Z(2), Z(2)], [0*Z(2), Z(2)]] ] ); ARepByImages( GroupWithGenerators( [ (1,2) ] ), [ [ [ Z(2)^0, Z(2)^0 ], [ 0*Z(2), Z(2)^0 ] ] ], "hom" ) gap> R.name := "R"; "R" gap> InductionARep(R, G); InductionARep( R, GroupWithGenerators( [ (1,2,3,4), (1,2) ] ), [ (), (3,4), (2,3), (2,3,4), (2,4,3), (2,4), (1,4,3), (1,4), (1,4,2,3), (1,4)(2,3), (1,2,3), (1,2,3,4) ] )
ExtensionARep( R, chi )
Let R be an irreducible arep of characteristic zero and chi a
character of a supergroup of R.source
which extends the
character of R. ExtensionARep
returns an arep of type
"extension"
representing an extension of R to
chi.source
. Here, ``supergroup" means
that all elements of R.source
are contained in G.
The extension is evaluated using Minkwitz's formula
(see Min96).
Note that R and chi can be accessed in the fields
.rep
and .character
of the result.
gap> G := GroupWithGenerators( [ (1,2,3,4), (1,2) ] ); Group( (1,2,3,4), (1,2) ) gap> H := GroupWithGenerators(AlternatingGroup(4)); Group( (1,2,4), (2,3,4) ) gap> G.name := "S4"; "S4" gap> H.name := "A4"; "A4" gap> R := ARepByImages(H, [ Mon( (1,2,3), [ 1, -1, -1 ] ), > Mon( (1,2,3), 3 ) ] ); ARepByImages( A4, [ Mon( (1,2,3), [ 1, -1, -1 ] ), Mon( (1,2,3), 3 ) ], "hom" ) gap> L := Irr(G); [ Character( Group( (1,2,3,4), (1,2) ), [ 1, 1, 1, 1, 1 ] ), Character( Group( (1,2,3,4), (1,2) ), [ 1, -1, 1, 1, -1 ] ), Character( Group( (1,2,3,4), (1,2) ), [ 2, 0, -1, 2, 0 ] ), Character( Group( (1,2,3,4), (1,2) ), [ 3, -1, 0, -1, 1 ] ), Character( Group( (1,2,3,4), (1,2) ), [ 3, 1, 0, -1, -1 ] ) ] gap> ExtensionARep(R, L[4]); ExtensionARep( ARepByImages( A4, [ Mon( (1,2,3), [ 1, -1, -1 ] ), Mon( (1,2,3), 3 ) ], "hom" ), Character( Group( (1,2,3,4), (1,2) ), [ 3, -1, 0, -1, 1 ] ) )
GaloisConjugateARep( R, aut )
GaloisConjugateARep( R, k )
GaloisConjugateARep
returns an arep of type "galoisConjugate"
representing the Galois conjugate of the arep A.
The conjugating automorphism may either be a field automorphism aut
or an integer k specifying the automorphism x -> GaloisCyc(x, k)
in the case characteristic = 0 or x -> x^(FrobeniusAut^k)
in the case characteristic = p prime.
The Galois conjugate of a representation R with a field automorphism aut is defined by x→ R(x)aut .
Note that R and aut can be accessed
in the fields .rep
and .galoisAut
resp. of the result.
gap> G := GroupWithGenerators( [ (1,2,3) ] ); Group( (1,2,3) ) gap> R := ARepByImages(G, [ [[E(3)]] ] ); ARepByImages( GroupWithGenerators( [ (1,2,3) ] ), [ [ [ E(3) ] ] ], "hom" ) gap> GaloisConjugateARep(R, -1); GaloisConjugateARep( ARepByImages( GroupWithGenerators( [ (1,2,3) ] ), [ [ [ E(3) ] ] ], "hom" ), -1 )
74.84 Basic Functions for AReps
The following sections describe basic functions for areps like e.g.
testing irreducibility and equivalence, evaluating an arep at a
group element, computing kernel and character, and constructing an arep
with given character.
74.85 Comparison of AReps
R1 = R2
R1 <> R2
The equality operator =
evaluates to true
if the areps R1
and R2 are equal and to false
otherwise. The inequality operator
<>
evaluates to true
if the amats R1 and R2 are not equal
and to false
otherwise.
Two areps are equal iff they define the same representation.
This means that first the sources have to be equal, i.e.
R1.source = R2.source
and second the images are
pointwise equal.
R1 < R2
R1 <= R2
R1 >= R2
R1 > R2
The operators <
, <=
, >=
, and >
evaluate to true
if the
arep R1 is strictly less than, less than or equal to, greater than or
equal to, and strictly greater than the arep R2.
The ordering of areps is defined via the ordering of records.
ImageARep( x, R )
or x ^ R
Let R be an arep and x a group element of R.source
.
ImageARep
returns the image of x under R as an amat
(see AMats). For conversion of amats see PermAMat -- MatAMat.
gap> G := GroupWithGenerators(SolvableGroup(8, 5)); Q8 gap> R := RegularARep(G); RegularARep( Q8 ) gap> x := Random(G); c gap> ImageARep(x, R); TensorProductAMat( AMatPerm((1,2)(3,4)(5,6)(7,8), 8), IdentityPermAMat(1) ) * DirectSumAMat( IdentityPermAMat(1), IdentityPermAMat(1), IdentityPermAMat(1), IdentityPermAMat(1), IdentityPermAMat(1), IdentityPermAMat(1), IdentityPermAMat(1), IdentityPermAMat(1) ) gap> PermAMat(last); (1,2)(3,4)(5,6)(7,8)
ImageARep( list, R )
ImageARep
returns the list of images of the group elements in
list under the arep R (see above). The images are amats
(see AMats). For conversion of amats see PermAMat -- MatAMat.
IsEquivalentARep( R1, R2 )
Let R1 and R2 be two areps with Maschke condition,
i.e. Size( Ri.source ) mod Ri.char
≠ 0, i = 1,2.
IsEquivalentARep
returns true
if the areps R1 and R2
define equivalent representations and false
otherwise.
Two representations (with Maschke condition) are equivalent iff they
have the same character.
R1 and R2 must have identical source
(i.e. IsIdentical(R1, R2) = true
)
and characteristic otherwise an error is signaled.
gap> G := GroupWithGenerators( [ (1,2,3) ] ); Group( (1,2,3) ) gap> R1 := NaturalARep(G, 3); NaturalARep( GroupWithGenerators( [ (1,2,3) ] ), 3 ) gap> R2 := RegularARep(G); RegularARep( GroupWithGenerators( [ (1,2,3) ] ) ) gap> IsEquivalentARep(R1, R2); true
CharacterARep( R )
CharacterARep
returns the character of the arep R.
Since GAP3 only provides characters of characteristic
zero, CharacterARep
only works in this case and will
signal an error otherwise.
Note that CharacterARep
sets and tests R.character
.
gap> G := GroupWithGenerators( [ (1,2), (3,4) ] ); Group( (1,2), (3,4) ) gap> CharacterARep(RegularARep(G)); Character( Group( (1,2), (3,4) ), [ 4, 0, 0, 0 ] )
IsIrreducibleARep( R )
Let R an arep of characteristic zero. IsIrreducibleARep
returns
true
if R represents an irreducible arep and false
otherwise. To determine irreducibility the character is used,
which is the reason for the condition characteristic = 0
(see CharacterARep).
Note that IsIrreducibleARep
sets and tests R.isIrreducible
.
gap> G := GroupWithGenerators(SolvableGroup(12, 5)); A4 gap> L := Irr(G); [ Character( A4, [ 1, 1, 1, 1 ] ), Character( A4, [ 1, 1, E(3), E(3)^2 ] ), Character( A4, [ 1, 1, E(3)^2, E(3) ] ), Character( A4, [ 3, -1, 0, 0 ] ) ] gap> R := ARepByCharacter(L[2]); ARepByImages( A4, [ Mon( [ E(3) ] ), Mon( (), 1 ), Mon( (), 1 ) ], "hom" ) gap> IsIrreducibleARep(R); true gap> IsIrreducibleARep(RegularARep(G)); false
KernelARep( R )
KernelARep
returns the kernel of the arep R.
Note that KernelARep
sets and tests R.kernel
.
gap> G := GroupWithGenerators(SymmetricGroup(3)); Group( (1,3), (2,3) ) gap> R := ARepByImages(G, [ [[-1]], [[-1]] ] ); ARepByImages( GroupWithGenerators( [ (1,3), (2,3) ] ), [ [ [ -1 ] ], [ [ -1 ] ] ], "hom" ) gap> KernelARep(R); Subgroup( Group( (1,3), (2,3) ), [ (1,3,2) ] )
IsFaithfulARep( R )
IsFaithfulARep
returns true
if the arep R represents a faithful
representation and false
otherwise.
Note that IsFaithfulARep
sets and tests R.isFaithful
.
gap> G := GroupWithGenerators(SolvableGroup(16, 7)); Q8x2 gap> IsFaithfulARep(TrivialPermARep(G)); false gap> IsFaithfulARep(RegularARep(G)); true
ARepWithCharacter( chi )
ARepWithCharacter
constructs an arep with character chi.
The group chi.source
must be solvable otherwise an error is
signaled. Note that the function returns a monomial arep
if this is possible.
Attention: ARepWithCharacter
only works in GAP3 3.4.4 after
bugfix 9!
gap> G := GroupWithGenerators(SolvableGroup(8, 5)); Q8 gap> L := Irr(G); [ Character( Q8, [ 1, 1, 1, 1, 1 ] ), Character( Q8, [ 1, 1, -1, 1, -1 ] ), Character( Q8, [ 1, 1, 1, -1, -1 ] ), Character( Q8, [ 1, 1, -1, -1, 1 ] ), Character( Q8, [ 2, -2, 0, 0, 0 ] ) ] gap> MonARepARep(ARepWithCharacter(L[5])); ARepByImages( Q8, [ Mon( (1,2), [ -1, 1 ] ), Mon( [ E(4), -E(4) ] ), Mon( [ -1, -1 ] ) ], "hom" )
GeneralFourierTransform( G )
GeneralFourierTransform
returns an amat representing
a Fourier transform over the complex numbers for the
solvable group G. For an explanation of Fourier transforms
see CB93. In order to obtain a fast Fourier transform
for G apply the function DecompositionMonRep
to any
regular representation of G.
Attention: GeneralFourierTransform
only works in GAP3 3.4.4 after
bugfix 9!
gap> G := SymmetricGroup(3); Group( (1,3), (2,3) ) gap> GeneralFourierTransform(G); AMatMat( [ [ 1, 1, 1, 1, 1, 1 ], [ 1, -1, -1, 1, 1, -1 ], [ 1, 0, 0, E(3), E(3)^2, 0 ], [ 0, 1, E(3)^2, 0, 0, E(3) ], [ 0, 1, E(3), 0, 0, E(3)^2 ], [ 1, 0, 0, E(3)^2, E(3), 0 ] ], "invertible" ) ^ -1
The following sections describe functions for convertibility and
conversion of arbitrary areps to areps of type
"perm"
, "mon"
, and "mat"
. As in AMat (see AMats)
the naming of the functions follows the usual GAP3-convention:
ChalkCheese
makes chalk from cheese. The parts in the name
(chalk, cheese) are:
IsPermRep( R )
IsPermRep
returns true
if R represents a permutation
representation and false
otherwise. Note that the name of
this function is not IsPermARep
since R can be an arep
of any type but represents a permutation representation in
the mathematical sense (every image is a permutation matrix).
Note that IsPermRep
sets and tests R.isPermRep
.
gap> G := GroupWithGenerators( [ (1,2) ] ); Group( (1,2) ) gap> R := ARepByImages(G, [ Mon( [1, -1] ) ] ); ARepByImages( GroupWithGenerators( [ (1,2) ] ), [ Mon( [ 1, -1 ] ) ], "hom" ) gap> IsPermRep(ConjugateARep(R, DFTAMat(2))); true
IsMonRep( R )
IsMonRep
returns true
if R represents a monomial
representation and false
otherwise. Note that the name of
this function is not IsMonARep
since R can be an arep
of any type but represents a monomial representation in
the mathematical sense (every image is a monomial matrix).
Note that IsMonRep
sets and tests R.isMonRep
.
gap> G := GroupWithGenerators(SolvableGroup(8, 5)); Q8 gap> R := RegularARep(G); RegularARep( Q8 ) gap> IsMonRep(InnerTensorProductARep(R, R)); true
PermARepARep( R )
PermARepARep
returns an arep of type "perm"
representing
the same representation as the arep R if possible. Otherwise
false
is returned.
Note that PermARepARep
sets and tests R.permARep
.
gap> G := GroupWithGenerators( [ (1,2) ] ); Group( (1,2) ) gap> R := ARepByImages(G, [ Mon( [1, -1] ) ] ); ARepByImages( GroupWithGenerators( [ (1,2) ] ), [ Mon( [ 1, -1 ] ) ], "hom" ) gap> PermARepARep(ConjugateARep(R, DFTAMat(2))); NaturalARep( GroupWithGenerators( [ (1,2) ] ), 2 ) gap> PermARepARep(R); false
MonARepARep( R )
MonARepARep
returns an arep of type "mon"
representing
the same representation as the arep R if possible. Otherwise
false
is returned.
Note that MonARepARep
sets and tests R.monARep
.
gap> G := GroupWithGenerators( [ (1,2,3), (1,2) ] ); Group( (1,2,3), (1,2) ) gap> R1 := ARepByImages(G, [ [[1]], [[-1]] ] ); ARepByImages( GroupWithGenerators( [ (1,2,3), (1,2) ] ), [ [ [ 1 ] ], [ [ -1 ] ] ], "hom" ) gap> R2 := NaturalARep(G, 4); NaturalARep( GroupWithGenerators( [ (1,2,3), (1,2) ] ), 4 ) gap> MonARepARep(InnerTensorProductARep(R1, R2)); ARepByImages( GroupWithGenerators( [ (1,2,3), (1,2) ] ), [ Mon( (1,2,3), 4 ), Mon( (1,2), [ -1, -1, -1, -1 ] ) ], "hom" )
MatARepARep( R )
MatARepARep
returns an arep of type "mat"
representing
the same representation as the arep R.
Note that MatARepARep
sets and tests R.matARep
.
gap> G := GroupWithGenerators( [ (1,2), (3,4) ] ); Group( (1,2), (3,4) ) gap> MatARepARep(RegularARep(G, 3)); ARepByImages( GroupWithGenerators( [ (1,2), (3,4) ] ), [ [ [ 0*Z(3), 0*Z(3), Z(3)^0, 0*Z(3) ], [ 0*Z(3), 0*Z(3), 0*Z(3), Z(3)^0 ], [ Z(3)^0, 0*Z(3), 0*Z(3), 0*Z(3) ], [ 0*Z(3), Z(3)^0, 0*Z(3), 0*Z(3) ] ], [ [ 0*Z(3), Z(3)^0, 0*Z(3), 0*Z(3) ], [ Z(3)^0, 0*Z(3), 0*Z(3), 0*Z(3) ], [ 0*Z(3), 0*Z(3), 0*Z(3), Z(3)^0 ], [ 0*Z(3), 0*Z(3), Z(3)^0, 0*Z(3) ] ] ], "hom" )
74.100 Higher Functions for AReps
The following sections describe functions allowing the
structural manipulation of, mainly monomial, areps. The idea is
to convert a given arep into a mathematical equal (not only equivalent!)
arep having more structure.
Examples are: converting a transitive monomial arep into
a conjugated induction (see TransitiveToInductionMonRep),
converting an induction into a conjugated
double induction (see InsertedInductionARep),
changing the transversal of an induction
(see TransversalChangeInductionARep),
decomposing a transitive monomial arep into a conjugated
outer tensor product
(see OuterTensorProductDecompositionMonRep)
and last but not least decomposing a monomial arep into a
conjugated sum of irreducibles
(see DecompositionMonRep).
The latter is one of the most interesting functions
of the package \AREP.
74.101 IsRestrictedCharacter
IsRestrictedCharacter( chi, chisub )
IsRestrictedCharacter
returns true
if the character chisub
is a restriction of the character chi to chisub.source
and
false
otherwise. All elements of chisub.source
must be
contained in chi.source
otherwise an error is signaled.
gap> G := SymmetricGroup(3); G.name := "S3"; Group( (1,3), (2,3) ) "S3" gap> H := CyclicGroup(3); H.name := "Z3"; Group( (1,2,3) ) "Z3" gap> L1 := Irr(G); [ Character( S3, [ 1, 1, 1 ] ), Character( S3, [ 1, -1, 1 ] ), Character( S3, [ 2, 0, -1 ] ) ] gap> L2 := Irr(H); [ Character( Z3, [ 1, 1, 1 ] ), Character( Z3, [ 1, E(3), E(3)^2 ] ), Character( Z3, [ 1, E(3)^2, E(3) ] ) ] gap> IsRestrictedCharacter(L1[2], L2[1]); true
AllExtendingCharacters( chi, G )
AllExtendingCharacters
returns the list of all characters
of G extending chi. All elements of chi.source
must be
contained in G otherwise an error is signaled.
gap> H := AlternatingGroup(4); H.name := "A4"; Group( (1,2,4), (2,3,4) ) "A4" gap> G := SymmetricGroup(4); G.name := "S4"; Group( (1,4), (2,4), (3,4) ) "S4" gap> L := Irr(H); [ Character( A4, [ 1, 1, 1, 1 ] ), Character( A4, [ 1, 1, E(3)^2, E(3) ] ), Character( A4, [ 1, 1, E(3), E(3)^2 ] ), Character( A4, [ 3, -1, 0, 0 ] ) ] gap> AllExtendingCharacters(L[4], G); [ Character( S4, [ 3, -1, -1, 0, 1 ] ), Character( S4, [ 3, 1, -1, 0, -1 ] ) ]
OneExtendingCharacter( chi, G )
OneExtendingCharacter
returns one character
of G extending chi if possible or returns false otherwise.
All elements of chi.source
must be
contained in G otherwise an error is signaled.
gap> H := Group( (1,3)(2,4) ); H.name := "Z2"; Group( (1,3)(2,4) ) "Z2" gap> G := Group( (1,2,3,4) ); G.name := "Z4"; Group( (1,2,3,4) ) "Z4" gap> L := Irr(H); [ Character( Z2, [ 1, 1 ] ), Character( Z2, [ 1, -1 ] ) ] gap> OneExtendingCharacter(L[2], G); Character( Z4, [ 1, E(4), -1, -E(4) ] )
IntertwiningSpaceARep( R1, R2 )
IntertwiningSpaceARep
returns a list of amats (see AMats)
representing a base of the intertwining space Int(R_1, R_2)
of the areps R1 and R2, which must have common source
and characteristic otherwise an error is signaled.
The intertwining space Int(R_1, R_2) of two representations R1 and R2 of a group G of the same characteristic is the vector space of matrices {M| R1(x). M = M . R2(x), for all x∈ G}.
gap> G := GroupWithGenerators( [ (1,2,3) ] ); Group( (1,2,3) ) gap> R1 := NaturalARep(G, 3); NaturalARep( GroupWithGenerators( [ (1,2,3) ] ), 3 ) gap> R2 := ARepByImages(G, [ Mon( [1, E(3), E(3)^2] ) ] ); ARepByImages( GroupWithGenerators( [ (1,2,3) ] ), [ Mon( [ 1, E(3), E(3)^2 ] ) ], "hom" ) gap> IntertwiningSpaceARep(R1, R2); [ AMatMat( [ [ 1, 0, 0 ], [ 1, 0, 0 ], [ 1, 0, 0 ] ] ), AMatMat( [ [ 0, 1, 0 ], [ 0, E(3), 0 ], [ 0, E(3)^2, 0 ] ] ), AMatMat( [ [ 0, 0, 1 ], [ 0, 0, E(3)^2 ], [ 0, 0, E(3) ] ] ) ]
IntertwiningNumberARep( R1, R2 )
IntertwiningNumberARep
returns the intertwining number of the
areps R1 and R2. The Maschke condition must hold for both
R1 and R2, otherwise an error is signaled.
R1 and R2 must have identical source
(i.e. IsIdentical(R1, R2) = true
)
and characteristic otherwise an error is signaled.
The intertwining number of two representations R1 and R2 (with Maschke condition) is the dimension of the intertwining space or the scalar product of the characters.
gap> G := GroupWithGenerators(SolvableGroup(64, 12)); 2^3xD8 gap> R := RegularARep(G); RegularARep( 2^3xD8 ) gap> IntertwiningNumberARep(R, R); 64
UnderlyingPermRep( R )
Let R be a monomial arep (i.e. IsMonRep( R ) = true
).
UnderlyingPermRep
returns an arep of type "perm"
representing the underlying permutation representation of R.
The underlying permutation representation of a monomial representation R is obtained by replacing all entries ≠ 0 in the images <R>(x), x∈ G by 1.
gap> G := GroupWithGenerators( [ (1,2) ] ); Group( (1,2) ) gap> R := ARepByImages(G, [ [[0, 2], [1/2, 0]] ] ); ARepByImages( GroupWithGenerators( [ (1,2) ] ), [ [ [ 0, 2 ], [ 1/2, 0 ] ] ], "hom" ) gap> UnderlyingPermARep(R); NaturalARep( GroupWithGenerators( [ (1,2) ] ), 2 )
IsTransitiveMonRep( R )
Let R be a monomial arep (i.e. IsMonRep( R ) = true
).
IsTransitiveMonRep
returns true
if R is transitive
and false
otherwise.
Note that IsTransitiveMonRep
sets and tests R.isTransitive
.
A monomial representation is transitive iff the underlying permutation representation is.
gap> G := GroupWithGenerators( [ (1,2), (3,4) ] ); Group( (1,2), (3,4) ) gap> IsTransitiveMonRep(NaturalARep(G, 4)); false gap> IsTransitiveMonRep(RegularARep(G)); true
IsPrimitiveMonRep( R )
Let R be a monomial arep (i.e. IsMonRep( R ) = true
).
IsPrimitiveMonRep
returns true
if R is primitive
and false
otherwise.
A monomial representation is primitive iff the underlying permutation representation is.
gap> G := GroupWithGenerators(SymmetricGroup(4)); G.name := "S4"; Group( (1,4), (2,4), (3,4) ) "S4" gap> H := GroupWithGenerators(SymmetricGroup(3)); H.name := "S3"; Group( (1,3), (2,3) ) "S3" gap> L := Irr(H); [ Character( S3, [ 1, 1, 1 ] ), Character( S3, [ 1, -1, 1 ] ), Character( S3, [ 2, 0, -1 ] ) ] gap> R := ARepByCharacter(L[2]); ARepByImages( S3, [ Mon( [ -1 ] ), Mon( [ -1 ] ) ], "hom" ) gap> IsPrimitiveMonRep(InductionARep(R, G)); true
74.109 TransitivityDegreeMonRep
TransitivityDegreeMonRep( R )
Let R be a monomial arep (i.e. IsMonRep( R ) = true
).
TransitivityDegreeMonRep
returns the degree of transitivity
of R as an integer.
Note that TransitivityDegreeMonRep
sets and tests R.transitivity
.
The degree of transitivity of a monomial representation is defined as the degree of transitivity of the underlying permutation representation.
gap> G := GroupWithGenerators(AlternatingGroup(5)); Group( (1,2,5), (2,3,5), (3,4,5) ) gap> TransitivityDegreeMonRep(NaturalARep(G, 5)); 3
74.110 OrbitDecompositionMonRep
OrbitDecompositionMonRep( R )
Let R be a monomial arep (i.e. IsMonRep( R ) = true
).
OrbitDecompositionMonRep
returns an arep equal to R with
structure (R1⊕...⊕ Rk)P
where Ri, i = 1,...,k are transitive areps of type "mon"
and
P is an amat of type "perm"
(for amats see AMats).
gap> G := GroupWithGenerators( [ (1,2,3,4) ] ); G.name := "Z4"; Group( (1,2,3,4) ) "Z4" gap> R := ARepByImages(G, [ Mon( (1,2)(3,4), [1,-1,1,1,-1] ) ] ); ARepByImages( GroupWithGenerators( [ (1,2,3,4) ] ), [ Mon( (1,2)(3,4), [ 1, -1, 1, 1, -1 ] ) ], "hom" ) gap> OrbitDecompositionMonRep(R); ConjugateARep( DirectSumARep( ARepByImages( Z4, [ Mon( (1,2), [ 1, -1 ] ) ], "hom" ), ARepByImages( Z4, [ Mon( (1,2), 2 ) ], "hom" ), ARepByImages( Z4, [ Mon( [ -1 ] ) ], "hom" ) ), IdentityPermAMat(5) )
74.111 TransitiveToInductionMonRep
TransitiveToInductionMonRep( R )
TransitiveToInductionMonRep( R, i )
Let R be a transitive monomial arep of a group G.
TransitiveToInductionMonRep
returns an arep equal to R
with structure <R> = (L\uparrowT G)D.
L is an arep of degree one of
the stabilizer H of the point i and T a transversal
of <H> \ G. The default for i is R.degree
.
D is a diagonal amat (see AMats) of type "mon"
.
Note that TransitiveToInductionMonRep
sets and tests the
field R.induction
if i = R.degree
.
gap> G := GroupWithGenerators(DihedralGroup(8)); Group( (1,2,3,4), (2,4) ) gap> R := ARepByImages(G, [ Mon( [E(4), E(4)^-1] ), Mon( (1,2), 2 ) ]); ARepByImages( GroupWithGenerators( [ (1,2,3,4), (2,4) ] ), [ Mon( [ E(4), -E(4) ] ), Mon( (1,2), 2 ) ], "hom" ) gap> TransitiveToInductionMonRep(R); ConjugateARep( InductionARep( ARepByImages( GroupWithGenerators( [ (1,2,3,4) ] ), [ Mon( [ -E(4) ] ) ], "hom" ), GroupWithGenerators( [ (1,2,3,4), (2,4) ] ), [ (2,4), () ] ), IdentityMonAMat(2) )
InsertedInductionARep( R, H )
Let R be an arep of type "induction"
, i.e.
<R> = L\uparrowT G where L is an arep of <U> ≤ G
and <U> ≤ H ≤ G.
InsertedInductionARep
returns an arep equal to R
with structure ( (L\uparrowT1H)\uparrowT2G)M
where M is an amat (see AMats) with a structure similar to R.
If R.rep
is of degree 1 then M is an amat of type "mon"
.
gap> G := GroupWithGenerators(SymmetricGroup(4)); G.name := "S4"; Group( (1,4), (2,4), (3,4) ) "S4" gap> H := GroupWithGenerators(AlternatingGroup(4)); H.name := "A4"; Group( (1,2,4), (2,3,4) ) "A4" gap> U := GroupWithGenerators(CyclicGroup(3)); U.name := "Z3"; Group( (1,2,3) ) "Z3" gap> R := ARepByImages(U, [ [[E(3)]] ] ); ARepByImages( Z3, [ [ [ E(3) ] ] ], "hom" ) gap> InsertedInductionARep(InductionARep(R, G), H); ConjugateARep( InductionARep( InductionARep( ARepByImages( Z3, [ [ [ E(3) ] ] ], "hom" ), A4, [ (), (2,3,4), (2,4,3), (1,4)(2,3) ] ), S4, [ (), (3,4) ] ), AMatMon( Mon( (2,4,8,7,3,5), [ 1, 1, 1, 1, 1, 1, E(3)^2, 1 ] ) ) )
ConjugationPermReps( R1, R2 )
Let R1 and R2 be permutation representations (i.e.
IsPermRep( Ri ) = true
, i = 1,2).
ConjugationPermReps
returns an amat A
(see AMats) of type "perm"
such that <R1>A = R_2.
R1 and R2 must have common source and characteristic
otherwise an error is signaled.
gap> G := GroupWithGenerators( [ (1,2,3) ] ); Group( (1,2,3) ) gap> R1 := NaturalARep(G, 3); NaturalARep( GroupWithGenerators( [ (1,2,3) ] ), 3 ) gap> R2 := ARepByImages(G, [ (1,3,2) ], 3); ARepByImages( GroupWithGenerators( [ (1,2,3) ] ), [ (1,3,2) ], 3, # degree "hom" ) gap> A := ConjugationPermReps(R1, R2); AMatPerm((2,3), 3) gap> R1 ^ A = R2; true
74.114 ConjugationTransitiveMonReps
ConjugationTransitiveMonReps( R1, R2 )
Let R1 and R2 be transitive monomial representations.
ConjugationTransitiveMonReps
returns an amat A
(see AMats) of type "mon"
such that <R1>A = R_2
if possible and false
otherwise.
R1 and R2 must have common source
otherwise an error is signaled.
Note that a conjugating monomial matrix exists iff R1 and R2 are induced from inner conjugated representations of degree one (see Pue98).
gap> G := GroupWithGenerators( [ (1,2,3), (1,2) ] ); Group( (1,2,3), (1,2) ) gap> R1 := ARepByImages(G, [ Mon( [E(3), E(3)^2] ), Mon( (1,2), 2 ) ]); ARepByImages( GroupWithGenerators( [ (1,2,3), (1,2) ] ), [ Mon( [ E(3), E(3)^2 ] ), Mon( (1,2), 2 ) ], "hom" ) gap> R2 := ARepByImages(G, [ Mon( [E(3)^2, E(3)] ), Mon( (1,2), 2 ) ]); ARepByImages( GroupWithGenerators( [ (1,2,3), (1,2) ] ), [ Mon( [ E(3)^2, E(3) ] ), Mon( (1,2), 2 ) ], "hom" ) gap> ConjugationTransitiveMonReps(R1, R2); AMatMon( Mon( (1,2), 2 ) )
74.115 TransversalChangeInductionARep
TransversalChangeInductionARep( R, T )
TransversalChangeInductionARep( R, T, hint )
Let R be an arep of type "induction"
, i.e.
<R> = L\uparrowS G and T another transversal of
L.source
\ G.
TransversalChangeInductionARep
returns an arep equal to R
with structure (L\uparrowT G)M
where M is an amat (see AMats). M is of type "mon"
if L is of degree 1 else M has a structure similar to R.
The hint "isTransversal"
suppresses checking T to be a
right transversal.
gap> G := GroupWithGenerators(SymmetricGroup(4)); G.name := "S4"; Group( (1,4), (2,4), (3,4) ) "S4" gap> H := GroupWithGenerators(SymmetricGroup(3)); H.name := "S3"; Group( (1,3), (2,3) ) "S3" gap> R := ARepByImages(H, [ [[-1]], [[-1]] ], "hom" ); ARepByImages( S3, [ [ [ -1 ] ], [ [ -1 ] ] ], "hom" ) gap> RG := InductionARep(R, G); InductionARep( ARepByImages( S3, [ [ [ -1 ] ], [ [ -1 ] ] ], "hom" ), S4, [ (), (3,4), (2,4), (1,4) ] ) gap> T := [(1,2,3,4), (2,3,4), (3,4), ()];; gap> TransversalChangeInductionARep(RG, T); ConjugateARep( InductionARep( ARepByImages( S3, [ [ [ -1 ] ], [ [ -1 ] ] ], "hom" ), S4, [ (1,2,3,4), (2,3,4), (3,4), () ] ), AMatMon( Mon( (1,4)(2,3), [ 1, 1, -1, 1 ] ) ) ) gap> last = RG; true
74.116 OuterTensorProductDecompositionMonRep
OuterTensorProductDecompositionMonRep( R )
Let R be a transitive monomial arep.
OuterTensorProductDecompositionMonRep
returns an arep
equal to R with structure (R_1# ...#R_k)M.
The Ri are areps
of type "mon"
, M is an amat of type mon.
For a definition of the outer tensor product of representations see OuterTensorProductARep. For an explanation of the algorithm see Pue98.
gap> G := GroupWithGenerators(SolvableGroup(48, 16)); 2x4xS3 gap> R := RegularARep(G, 2); RegularARep( 2x4xS3, GF(2) ) gap> OuterTensorProductDecompositionMonRep(R); ConjugateARep( OuterTensorProductARep( 2x4xS3, ARepByImages( GroupWithGenerators( [ c ] ), [ Mon( (1,2), 2, GF(2) ) ], "hom" ), ARepByImages( GroupWithGenerators( [ d, e ] ), [ Mon( (1,3,2,4), 4, GF(2) ), Mon( (1,2)(3,4), 4, GF(2) ) ], "hom" ), ARepByImages( GroupWithGenerators( [ a*e, b ] ), [ Mon( (1,4)(2,6)(3,5), 6, GF(2) ), Mon( (1,2,3)(4,5,6), 6, GF(2) ) ], "hom" ) ), AMatMon( Mon( ( 2, 9,18,44,16,28,30,46,31, 6,42,48,47,39,23,35,37, 7) ( 3,17,36,45,24,43, 8,10,25, 5,34,29,38,15,19, 4,26,13) (11,33,22,27,21,20,12,41,40,32,14), 48, GF(2) ) ) ) gap> last = R; true
InnerConjugationARep( R, G, t )
Let R be an arep with source <H> ≤ G and <t>∈ G.
InnerConjugationARep
returns an arep of type
"perm"
or "mon"
or "mat"
, the most specific
possible, representing the inner conjugate <R>t
of R with t.
The inner conjugate <R>t is a representation of <H>t defined by x→ R(t. x. t-1).
gap> G := GroupWithGenerators(SymmetricGroup(4)); Group( (1,4), (2,4), (3,4) ) gap> H := GroupWithGenerators(SymmetricGroup(3)); Group( (1,3), (2,3) ) gap> R := NaturalARep(H, 3); NaturalARep( GroupWithGenerators( [ (1,3), (2,3) ] ), 3 ) gap> InnerConjugationARep(R, G, (1,2,3,4)); ARepByImages( GroupWithGenerators( [ (2,4), (3,4) ] ), [ (1,3), (2,3) ], 3, # degree "hom" )
74.118 RestrictionInductionARep
RestrictionInductionARep( R, K )
Let R be an arep of type "induction"
, i.e.
<R> = L\uparrowT G where L is an arep of <H> ≤ G of
degree 1 and <K> ≤ G a subgroup.
RestrictionInductionARep
returns an arep equal to the
restriction <R>\downarrow K with structure
(⊕i=1k
((Lsi\downarrow(Hsi∩ K))\uparrowTi K))M.
S = {s1,..., sk} is a transversal of the double cosets
H \ G/K, <L>si denotes the inner conjugate of R
with si, and M is an amat (see AMats) of type "mon"
.
Note that this decomposition is based on a refined version of Mackey's subgroup theorem (see Pue98).
gap> G := GroupWithGenerators(SymmetricGroup(4)); G.name := "S4"; Group( (1,4), (2,4), (3,4) ) "S4" gap> H := GroupWithGenerators( [ (1,2) ] ); H.name := "Z2"; Group( (1,2) ) "Z2" gap> K := GroupWithGenerators( [ (1,2,3) ] ); K.name := "Z3"; Group( (1,2,3) ) "Z3" gap> L := ARepByImages(H, [ Mon( [-1] ) ] ); ARepByImages( Z2, [ Mon( [ -1 ] ) ], "hom" ) gap> RestrictionInductionARep(InductionARep(L, G), K); ConjugateARep( DirectSumARep( RegularARep( GroupWithGenerators( [ (1,2,3) ] ) ), RegularARep( GroupWithGenerators( [ (1,2,3) ] ) ), RegularARep( GroupWithGenerators( [ (1,2,3) ] ) ), RegularARep( GroupWithGenerators( [ (1,2,3) ] ) ) ), AMatMon( Mon( ( 2,12, 4, 6, 9, 5, 8,10), [ 1, 1, -1, -1, 1, 1, -1, -1, -1, -1, 1, 1 ] ) ) )
kbsARep( R )
kbsARep
returns the kbs (conjugated blockstructure)
of the arep R. The kbs of a representation is a partition
of the set {1,...,R.degree
} representing the blockstructure
of R. For an exact definition see kbs.
Note that for a monomial representation the kbs is exactly the list of orbits.
gap> G := GroupWithGenerators( [ (1,2) ] ); Group( (1,2) ) gap> R := ARepByImages(G, [ (2,3) ], 4); ARepByImages( GroupWithGenerators( [ (1,2) ] ), [ (2,3) ], 4, # degree "hom" ) gap> kbsARep(R); [ [ 1 ], [ 2, 3 ], [ 4 ] ]
74.120 RestrictionToSubmoduleARep
RestrictionToSubmoduleARep( R, list )
RestrictionToSubmoduleARep( R, list, hint )
Let R be an arep and list a subset of [1..R.degree]
.
RestrictionToSubmoduleARep
returns an arep of type
"perm"
or "mon"
or "mat"
, the most specific
possible, representing the restriction of R to the submodule
generated by the base vectors given through list.
The optional hint "hom"
avoids the check for homomorphism.
Note that the restriction to the submodule given by list defines a homomorphism iff list is a union of lists in the kbs of R (see kbsARep).
gap> G := GroupWithGenerators( [ (1,2) ] ); Group( (1,2) ) gap> R := ARepByImages(G, [ (2,4) ], 4); ARepByImages( GroupWithGenerators( [ (1,2) ] ), [ (2,4) ], 4, # degree "hom" ) gap> RestrictionToSubmoduleARep(R, [2,4]); NaturalARep( GroupWithGenerators( [ (1,2) ] ), 2 )
kbsDecompositionARep( R )
kbsDecompositionARep
returns an arep equal to R with
structure (R_1⊕...⊕R_k)P where P is
an amat (see AMats) of type "perm"
and all Ri have
trivial kbs (see kbsARep).
Note that for a monomial arep kbsDecompositionARep
performs
exactly the same as the function OrbitDecompositionMonRep
(see OrbitDecompositionMonRep).
gap> G := GroupWithGenerators( [ (1,2) ] ); Group( (1,2) ) gap> R := ARepByImages(G, > [ [[Z(2), Z(2), 0*Z(2), 0*Z(2)], [0*Z(2), Z(2), 0*Z(2), 0*Z(2)], > [0*Z(2), 0*Z(2), Z(2), Z(2)], [0*Z(2), 0*Z(2), 0*Z(2), Z(2)]] ] ); ARepByImages( GroupWithGenerators( [ (1,2) ] ), [ [ [ Z(2)^0, Z(2)^0, 0*Z(2), 0*Z(2) ], [ 0*Z(2), Z(2)^0, 0*Z(2), 0*Z(2) ], [ 0*Z(2), 0*Z(2), Z(2)^0, Z(2)^0 ], [ 0*Z(2), 0*Z(2), 0*Z(2), Z(2)^0 ] ] ], "hom" ) gap> kbsDecompositionARep(R); ConjugateARep( DirectSumARep( ARepByImages( GroupWithGenerators( [ (1,2) ] ), [ [ [ Z(2)^0, Z(2)^0 ], [ 0*Z(2), Z(2)^0 ] ] ], "hom" ), ARepByImages( GroupWithGenerators( [ (1,2) ] ), [ [ [ Z(2)^0, Z(2)^0 ], [ 0*Z(2), Z(2)^0 ] ] ], "hom" ) ), IdentityPermAMat(4, GF(2)) )
74.122 ExtensionOnedimensionalAbelianRep
ExtensionOnedimensionalAbelianRep( R, G )
Let R be an arep of the subgroup <H> ≤ G and
let <G>/kernel(R) be an abelian factor group.
ExtensionOnedimensionalAbelianRep
returns an arep of type
"mon"
and degree 1 extending R to G.
For the extension the smallest possible extension field
is chosen.
gap> G := GroupWithGenerators(CyclicGroup(8)); Group( (1,2,3,4,5,6,7,8) ) gap> H := GroupWithGenerators( [ G.1^2 ] ); Group( (1,3,5,7)(2,4,6,8) ) gap> R := ARepByImages(H, [ [[-1]] ] ); ARepByImages( GroupWithGenerators( [ (1,3,5,7)(2,4,6,8) ] ), [ [ [ -1 ] ] ], "hom" ) gap> ExtensionOnedimensionalAbelianRep(R, G); ARepByImages( GroupWithGenerators( [ (1,2,3,4,5,6,7,8) ] ), [ Mon( [ E(4) ] ) ], "hom" )
DecompositionMonRep( R )
DecompositionMonRep( R, hint )
Let R be a monomial arep (i.e. IsMonRep( R )
yields true
).
DecompositionMonRep
returns an arep equal to R with
structure (R_1⊕...⊕ R_k)A-1 where all
Ri are irreducible and A-1 is a highly structured
amat (see AMats). A is a decomposition matrix for R and
can be accessed in the field .conjugation.element
of the result.
The list of the Ri can be accessed in the field
.rep.summands
of the result. Note that any Ri is monomial if
this is possible.
If the hint "noOuter"
is supplied, the decomposition
of R is performed without any decomposition into
an outer tensor product which may speed up the function.
The function only works for characteristic zero otherwise an error
is signaled.
At least the following types of monomial areps can be decomposed:
monomial representations of solvable groups, double transitive
permutation representations, primitive permutation representations
with solvable socle.
If DecompositionMonRep
is not able to decompose R then false
is
returned.
The performance of DecompositionMonRep
depends on the size of the
group represented as well as on the degree of R. E.g. the
decomposition of a regular representation of a group of size 96
takes less than half a minute
(CPU-time on a SUN Ultra-Sparc 150\mathrmMHz) if the
source group is an ag group.
Note that in the case that R is a regular representation of the
solvable group G the structured decomposition matrix A computed by
DecompositionMonRep
represents a fast Fourier transform for G.
Hence, DecompositionMonRep
is able to compute a fast Fourier
transform for any solvable group.
The algorithm is a major result of Pue98 where a thorough explanation can be found.
Set InfoLatticeDec := Print
to obtain information on
the recursive decomposition of R.
An important application of this function is the automatic generation of fast algorithms for discrete signal transforms which is realized in Matrix Decomposition. (see Min93, Egn97, Pue98).
gap> G := GroupWithGenerators(SolvableGroup(8, 5)); Q8 gap> R := RegularARep(G); RegularARep( Q8 ) gap> DecompositionMonRep(R); ConjugateARep( DirectSumARep( TrivialMonARep( Q8 ), ARepByImages( Q8, [ Mon( [ -1 ] ), Mon( [ -1 ] ), Mon( (), 1 ) ], "hom" ), ARepByImages( Q8, [ Mon( [ -1 ] ), Mon( (), 1 ), Mon( (), 1 ) ], "hom" ), ARepByImages( Q8, [ Mon( (), 1 ), Mon( [ -1 ] ), Mon( (), 1 ) ], "hom" ), ARepByImages( Q8, [ Mon( (1,2), [ -1, 1 ] ), Mon( [ E(4), -E(4) ] ), Mon( [ -1, -1 ] ) ], "hom" ), ARepByImages( Q8, [ Mon( (1,2), [ -1, 1 ] ), Mon( [ E(4), -E(4) ] ), Mon( [ -1, -1 ] ) ], "hom" ) ), ( AMatPerm((7,8), 8) * TensorProductAMat( IdentityPermAMat(2), AMatPerm((2,3), 4) * TensorProductAMat( DFTAMat(2), IdentityPermAMat(2) ) * DiagonalAMat([ 1, 1, 1, E(4) ]) * TensorProductAMat( IdentityPermAMat(2), DFTAMat(2) ) * AMatPerm((2,3), 4) ) * AMatMon( Mon( (2,5,3)(4,8,7), [ 1, 1, 1, 1, 1, 1, -1, 1 ] ) ) * DirectSumAMat( TensorProductAMat( DFTAMat(2), IdentityPermAMat(2) ), IdentityPermAMat(4) ) * AMatPerm((2,4), 8) ) ^ -1 ) gap> last = R; true
The following sections describe functions for the computation of symmetry of a given matrix. A symmetry of a matrix is a pair (R1, R2) of representations of the same group G with the property R1(x). M = M. R2(x) for all x∈ G. This definition corresponds to the definition of the intertwining space of R1, R2 (see IntertwiningSpaceARep). The origin of this definition is due to Minkwitz (see Min95, Min93) and was generalized to the definition above by the authors of this package.
Restrictions on the representations R1, R2 yield special types of symmetry. We consider the following three types:
• Perm-Irred symmetry: R1 is a permutation representation, R2 is a conjugated (by a permutation) direct sum of irreducible representations
• Perm-Perm symmetry: both R1 and R2 are permutation representations
• Mon-Mon symmetry: both R1 and R2 are monomial representations
There are two implementations for the search algorithm for
Perm-Perm-Symmetry. One is entirely in GAP3 by S. Egner, the other
uses the external C-program desauto
bei J. Leon which is distributed
with the \sf GUAVA package. By default the GAP3 code is run. In
order to use the much faster method of J. Leon based on partitions
(see Leo91) you should set UseLeon := true
and make sure
that an executable version of desauto
is placed in
\$GAP/pkg/arep/bin
. The implementation of Leon requires the matrix
to have ≤ 256 different entries. If this condition is violated
the GAP3 implementation is run.
A matrix with symmetry of one of the types above contains structure in a sense and can be decomposed into a product of highly structured sparse matrices (see Matrix Decomposition).
For details on the concept and computation of symmetry see Egn97 and Pue98.
The following functions are implemented in the file
"arep/lib/symmetry.g"
based on functions from
"arep/lib/permperm.g"
, "arep/lib/monmon.g"
,
"arep/lib/permblk.g"
and "arep/lib/permmat.g"
.
PermPermSymmetry( M )
Let M be a matrix or an amat (see AMats).
PermPermSymmetry
returns a pair (R1, R2) of
areps of type "perm"
(see AReps) of the same group G
representing the perm-perm symmetry of M, i.e.
R1(x). M = M. R2(x) for all x∈ G.
The returned symmetry is maximal in the sense that for every
pair (p1, p2) of permutations satisfying p1. M= M. p2
there is an x with p1 = R1(x) and p2 = R2(x).
To use the much faster implementation of J. Leon set
UseLeon := true
as explained in Symmetry of Matrices.
Set InfoPermSym1 := true
to obtain information
about the search.
For the algorithm see Leo91 resp. Egn97.
gap> M := DFT(5);; gap> PrintArray(M); [ [ 1, 1, 1, 1, 1 ], [ 1, E(5), E(5)^2, E(5)^3, E(5)^4 ], [ 1, E(5)^2, E(5)^4, E(5), E(5)^3 ], [ 1, E(5)^3, E(5), E(5)^4, E(5)^2 ], [ 1, E(5)^4, E(5)^3, E(5)^2, E(5) ] ] gap> L := PermPermSymmetry(M); [ ARepByImages( GroupWithGenerators( [ g1, g2 ] ), [ (2,3,5,4), (2,5)(3,4) ], 5, # degree "hom" ), ARepByImages( GroupWithGenerators( [ g1, g2 ] ), [ (2,4,5,3), (2,5)(3,4) ], 5, # degree "hom" ) ] gap> L[1]^AMatMat(M) = L[2]; true
MonMonSymmetry( M )
Let M be a matrix or an amat (see AMats) of characteristic zero.
MonMonSymmetry
returns a pair (R1, R2) of
areps of type "mon"
(see AReps) of the same group G
representing a mon-mon symmetry of M, i.e.
R1(x). M = M. R2(x) for all x∈ G.
The non-zero entries in the matrices R1(x), R2(x) are all roots of unity of a certain order d. This order is given by the lcm of all quotients of non-zero entries of M with equal absolute value. The returned symmetry is maximal in the sense that for every pair (m1, m2) of monomial matrices containing only dth roots of unity (and 0) and satisfying m1. M=M. m2 there is an x with m1 = R1(x) and m2 = R2(x).
MonMonSymmetry
uses the function PermPermSymmetry
. Hence you can
accelerate the function using the faster implementation of J. Leon
by setting UseLeon := true
as explained in Symmetry of Matrices.
For an explanation of the algorithm see Pue98.
gap> M := DFT(5);; gap> PrintArray(M); [ [ 1, 1, 1, 1, 1 ], [ 1, E(5), E(5)^2, E(5)^3, E(5)^4 ], [ 1, E(5)^2, E(5)^4, E(5), E(5)^3 ], [ 1, E(5)^3, E(5), E(5)^4, E(5)^2 ], [ 1, E(5)^4, E(5)^3, E(5)^2, E(5) ] ] gap> L := MonMonSymmetry(M); [ ARepByImages( GroupWithGenerators( [ g1, g2, g3, g4, g5 ] ), [ Mon( (2,3,5,4), [ 1, E(5)^3, E(5), E(5)^4, E(5)^2 ] ), Mon( (2,5)(3,4), [ 1, E(5)^2, E(5)^4, E(5), E(5)^3 ] ), Mon( (1,2,3,4,5), [ E(5), E(5), E(5), E(5), E(5) ] ), Mon( [ E(5), 1, E(5)^4, E(5)^3, E(5)^2 ] ), Mon( [ 1, E(5), E(5)^2, E(5)^3, E(5)^4 ] ) ], "hom" ), ARepByImages( GroupWithGenerators( [ g1, g2, g3, g4, g5 ] ), [ Mon( (1,3,4,2), 5 ), Mon( (1,4)(2,3), 5 ), Mon( [ E(5), E(5)^2, E(5)^3, E(5)^4, 1 ] ), Mon( (1,2,3,4,5), [ E(5), E(5), E(5), E(5), E(5) ] ), Mon( (1,5,4,3,2), 5 ) ], "hom" ) ] gap> L[1]^AMatMat(M) = L[2]; true
PermIrredSymmetry( M )
PermIrredSymmetry( M, maxblocksize )
Let M be a matrix or an amat (see AMats) of characteristic zero.
PermIrredSymmetry
returns a list of pairs (R1, R2) of
areps (see AReps) of the same group G
representing a perm-irred symmetry of M, i.e.
R1(x). M = M. R2(x) for all x∈ G and
R1 is a permutation representation and R2 a
conjugated (by a permutation) direct sum of irreducible
representations.
If maxblocksize is supplied exactly those perm-irred symmetries
are returned where R2 contains at least one irreducible of degree
≤ maxblocksize. The default for maxblocksize is 2.
Refer to Egn97 to understand how the search is done and how to interpret the result.
Note that the perm-irred symmetry is not symmetric. Hence it is possible that a matrix M admits a perm-irred symmetry but its transpose not.
The perm-irred symmetry is a special case of a perm-block symmetry.
The perm-block symmetries admitted by a fixed matrix M can be
described by two lattices which are in a certain way related to
each other (semi-order preserving). To explore this structure
(described in Egn97) you should refer to PermBlockSym
and DisplayPermBlockSym
in the file "arep/lib/permblk.g"
.
gap> M := DFT(4); [ [ 1, 1, 1, 1 ], [ 1, E(4), -1, -E(4) ], [ 1, -1, 1, -1 ], [ 1, -E(4), -1, E(4) ] ] gap> PermIrredSymmetry(M); [ [ NaturalARep( G2, 4 ), ConjugateARep( DirectSumARep( TrivialMatARep( G2 ), ARepByImages( G2, [ [ [ -1 ] ], [ [ E(4) ] ] ], "hom" ), ARepByImages( G2, [ [ [ 1 ] ], [ [ -1 ] ] ], "hom" ), ARepByImages( G2, [ [ [ -1 ] ], [ [ -E(4) ] ] ], "hom" ) ), IdentityPermAMat(4) ) ], [ NaturalARep( G3, 4 ), ConjugateARep( DirectSumARep( TrivialMatARep( G3 ), ARepByImages( G3, [ [ [ 0, -E(4) ], [ E(4), 0 ] ], [ [ 0, 1 ], [ 1, 0 ] ], [ [ 0, -1 ], [ -1, 0 ] ] ], "hom" ), ARepByImages( G3, [ [ [ -1 ] ], [ [ 1 ] ], [ [ 1 ] ] ], "hom" ) ), AMatPerm((3,4), 4) ) ], [ NaturalARep( G1, 4 ), ConjugateARep( DirectSumARep( TrivialMatARep( G1 ), ARepByImages( G1, [ [ [ 1/2, -1/2+1/2*E(4), 1/2*E(4) ], [ -1/2-1/2*E(4), 0, -1/2+1/2*E(4) ], [ -1/2*E(4), -1/2-1/2*E(4), 1/2 ] ], [ [ 0, 0, 1 ], [ 0, 1, 0 ], [ 1, 0, 0 ] ], [ [ 1/2, 1/2+1/2*E(4), -1/2*E(4) ], [ 1/2-1/2*E(4), 0, 1/2+1/2*E(4) ], [ 1/2*E(4), 1/2-1/2*E(4), 1/2 ] ] ], "hom" ) ), IdentityPermAMat(4) ) ] ]
74.128 Discrete Signal Transforms
The following sections describe functions for the construction of many well known signal transforms in matrix form, as e.g. the discrete Fourier transform, several discrete cosine transforms etc. For the definition of the mentioned signal transforms see ER82, Mal92, Mer96.
The functions for discrete signal transforms are implemented
in "arep/lib/transf.g"
.
74.129 DiscreteFourierTransform
DiscreteFourierTransform( r )
DiscreteFourierTransform( n )
DiscreteFourierTransform( n, char )
shortcut: DFT
DiscreteFourierTransform
or DFT
returns the discrete Fourier
transform from a given root of unity r or the size n and
the characteristic char (see CB93).
The default for char is zero.
Note that the DFT on n points and characteristic char exists
iff n and char are coprime. If this condition is violenced
an error is signaled.
The DFTn of size n is defined as DFTn = [ωnkl| k,l∈{0,...,n-1}], ωn a primitive nth root of unity.
gap> DFT(Z(3)); [ [ Z(3)^0, Z(3)^0 ], [ Z(3)^0, Z(3) ] ] gap> DFT(4); [ [ 1, 1, 1, 1 ], [ 1, E(4), -1, -E(4) ], [ 1, -1, 1, -1 ], [ 1, -E(4), -1, E(4) ] ]
74.130 InverseDiscreteFourierTransform
InverseDiscreteFourierTransform( r )
InverseDiscreteFourierTransform( n )
InverseDiscreteFourierTransform( n, char )
shortcut: InvDFT
InverseDiscreteFourierTransform
or InvDFT
returns the inverse
of the discrete Fourier transform from a given root of unity r
or the size n and the characteristic char
(see DiscreteFourierTransform). The default for char is zero.
gap> InvDFT(3); [ [ 1/3, 1/3, 1/3 ], [ 1/3, 1/3*E(3)^2, 1/3*E(3) ], [ 1/3, 1/3*E(3), 1/3*E(3)^2 ] ]
74.131 DiscreteHartleyTransform
DiscreteHartleyTransform( n )
shortcut: DHT
DiscreteHartleyTransform
or DHT
returns the discrete
Hartley transform on n points.
The DHTn of size n is defined by DHTn = [1/√n. (cos(2π kl/n) + sin(2π kl/n))| k,l∈{0,...,n-1}].
gap> DHT(4); [ [ 1/2, 1/2, 1/2, 1/2 ], [ 1/2, 1/2, -1/2, -1/2 ], [ 1/2, -1/2, 1/2, -1/2 ], [ 1/2, -1/2, -1/2, 1/2 ] ]
74.132 InverseDiscreteHartleyTransform
InverseDiscreteHartleyTransform( n )
shortcut: InvDHT
InverseDiscreteHartleyTransform
or InvDHT
returns the inverse
of the discrete Hartley transform on n points. Since the DHT is self
inverse the result is exactly the same as from DHT
above.
gap> InvDHT(4); [ [ 1/2, 1/2, 1/2, 1/2 ], [ 1/2, 1/2, -1/2, -1/2 ], [ 1/2, -1/2, 1/2, -1/2 ], [ 1/2, -1/2, -1/2, 1/2 ] ]
74.133 DiscreteCosineTransform
DiscreteCosineTransform( n )
shortcut: DCT
DiscreteCosineTransform
returns the standard cosine transform
(type II) on n points.
The DCTn of size n is defined by DCTn = [√2/n. ck. (cos(k(l+1/2)π/n)| k,l∈{0,...,n-1}], ck = 1/√2 for k = 0 and ck = 1 else.
gap> DCT(3); [ [ 1/3*E(12)^7-1/3*E(12)^11, 1/3*E(12)^7-1/3*E(12)^11, 1/3*E(12)^7-1/3*E(12)^11 ], [ -1/2*E(8)+1/2*E(8)^3, 0, 1/2*E(8)-1/2*E(8)^3 ], [ -1/6*E(24)+1/6*E(24)^11+1/6*E(24)^17-1/6*E(24)^19, 1/3*E(24)-1/3*E(24)^11-1/3*E(24)^17+1/3*E(24)^19, -1/6*E(24)+1/6*E(24)^11+1/6*E(24)^17-1/6*E(24)^19 ] ]
74.134 InverseDiscreteCosineTransform
InverseDiscreteCosineTransform( n )
shortcut: InvDCT
InverseDiscreteCosineTransform
returns the inverse of the
standard cosine transform (type II) on n points. Since the
DCT is orthogonal, the result is the transpose of the DCT,
which is exactly the discrete cosine transform of type III.
[ [ 1/3*E(12)^7-1/3*E(12)^11, -1/2*E(8)+1/2*E(8)^3, -1/6*E(24)+1/6*E(24)^11+1/6*E(24)^17-1/6*E(24)^19 ], [ 1/3*E(12)^7-1/3*E(12)^11, 0, 1/3*E(24)-1/3*E(24)^11-1/3*E(24)^17+1/3*E(24)^19 ], [ 1/3*E(12)^7-1/3*E(12)^11, 1/2*E(8)-1/2*E(8)^3, -1/6*E(24)+1/6*E(24)^11+1/6*E(24)^17-1/6*E(24)^19 ] ]
74.135 DiscreteCosineTransformIV
DiscreteCosineTransformIV( n )
shortcut: DCT_IV
DiscreteCosineTransformIV
returns the cosine transform of
type IV on n points.
The DCT\_IVn of size n is defined by DCT\_IVn = [√2/n. (cos((k+1/2)(l+1/2)π/n)| k,l∈{0,...,n-1}].
[ [ 1/2*E(12)^4+1/6*E(12)^7+1/2*E(12)^8-1/6*E(12)^11, 1/3*E(12)^7-1/3*E(12)^11, 1/2*E(12)^4-1/6*E(12)^7+1/2*E(12)^8+1/6*E(12)^11 ], [ 1/3*E(12)^7-1/3*E(12)^11, -1/3*E(12)^7+1/3*E(12)^11, -1/3*E(12)^7+1/3*E(12)^11 ], [ 1/2*E(12)^4-1/6*E(12)^7+1/2*E(12)^8+1/6*E(12)^11, -1/3*E(12)^7+1/3*E(12)^11, 1/2*E(12)^4+1/6*E(12)^7+1/2*E(12)^8-1/6*E(12)^11 ] ]
74.136 InverseDiscreteCosineTransformIV
InverseDiscreteCosineTransformIV( n )
shortcut: InvDCT_IV
InverseDiscreteCosineTransformIV
returns the inverse of the
cosine transform of type IV on n points. Since the
DCT\_IV is orthogonal, the result is the transpose of the DCT\_IV.
[ [ 1/3*E(12)^7-1/3*E(12)^11, -1/2*E(8)+1/2*E(8)^3, -1/6*E(24)+1/6*E(24)^11+1/6*E(24)^17-1/6*E(24)^19 ], [ 1/3*E(12)^7-1/3*E(12)^11, 0, 1/3*E(24)-1/3*E(24)^11-1/3*E(24)^17+1/3*E(24)^19 ], [ 1/3*E(12)^7-1/3*E(12)^11, 1/2*E(8)-1/2*E(8)^3, -1/6*E(24)+1/6*E(24)^11+1/6*E(24)^17-1/6*E(24)^19 ] ]
74.137 DiscreteCosineTransformI
DiscreteCosineTransformI( n )
shortcut: DCT_I
DiscreteCosineTransformI
returns the cosine transform of
type I on <n>+1 points.
The DCT\_In of size <n>+1 is defined by DCT\_In = [√2/n. ck. cl. (cos(klπ/n)| k,l∈{0,...,n}], ck = 1/√2 for k = 0 and ck = 1 else.
[ [ 1/2, 1/2*E(8)-1/2*E(8)^3, 1/2 ], [ 1/2*E(8)-1/2*E(8)^3, 0, -1/2*E(8)+1/2*E(8)^3 ], [ 1/2, -1/2*E(8)+1/2*E(8)^3, 1/2 ] ]
74.138 InverseDiscreteCosineTransformI
InverseDiscreteCosineTransformI( n )
shortcut: InvDCT_I
InverseDiscreteCosineTransformI
returns the inverse of the
cosine transform of type I on n points. Since the
DCT\_I is orthogonal, the result is the transpose of the DCT\_I.
[ [ 1/2, 1/2*E(8)-1/2*E(8)^3, 1/2 ], [ 1/2*E(8)-1/2*E(8)^3, 0, -1/2*E(8)+1/2*E(8)^3 ], [ 1/2, -1/2*E(8)+1/2*E(8)^3, 1/2 ] ]
WalshHadamardTransform( n )
shortcut: WHT
WalshHadamardTransform
returns the Walsh-Hadamard transform
on n points.
Let <n> = ∏i=1k piνi be the prime factor decomposition of n. Then the WHTn is defined by WHTn = ⊗i=1k DFTpi⊗νi.
gap> WHT(4); [ [ 1, 1, 1, 1 ], [ 1, -1, 1, -1 ], [ 1, 1, -1, -1 ], [ 1, -1, -1, 1 ] ]
74.140 InverseWalshHadamardTransform
InverseWalshHadamardTransform( n )
shortcut: InvWHT
InverseWalshHadamardTransform
returns the inverse of the
Walsh-Hadamard transform on n points.
gap> InvWHT(4); [ [ 1/4, 1/4, 1/4, 1/4 ], [ 1/4, -1/4, 1/4, -1/4 ], [ 1/4, 1/4, -1/4, -1/4 ], [ 1/4, -1/4, -1/4, 1/4 ] ]
SlantTransform( n )
shortcut: ST
SlantTransform
returns the Slant transform on n points, which
must be a power of 2, <n> = 2k
For a definition of the Slant transform see ER82, 10.9.
gap> ST(4); [ [ 1/2, 1/2, 1/2, 1/2 ], [ 3/10*E(5)-3/10*E(5)^2-3/10*E(5)^3+3/10*E(5)^4, 1/10*E(5)-1/10*E(5)^2-1/10*E(5)^3+1/10*E(5)^4, -1/10*E(5)+1/10*E(5)^2+1/10*E(5)^3-1/10*E(5)^4, -3/10*E(5)+3/10*E(5)^2+3/10*E(5)^3-3/10*E(5)^4 ], [ 1/2, -1/2, -1/2, 1/2 ], [ 1/10*E(5)-1/10*E(5)^2-1/10*E(5)^3+1/10*E(5)^4, -3/10*E(5)+3/10*E(5)^2+3/10*E(5)^3-3/10*E(5)^4, 3/10*E(5)-3/10*E(5)^2-3/10*E(5)^3+3/10*E(5)^4, -1/10*E(5)+1/10*E(5)^2+1/10*E(5)^3-1/10*E(5)^4 ] ]
InverseSlantTransform( n )
shortcut: InvST
InverseSlantTransform
returns the inverse of the
Slant transform on n points, which must be a power of 2, <n> = 2k.
Since ST is orthogonal, this is exactly the transpose
of the ST.
gap> InvST(4); [ [ 1/2, 3/10*E(5)-3/10*E(5)^2-3/10*E(5)^3+3/10*E(5)^4, 1/2, 1/10*E(5)-1/10*E(5)^2-1/10*E(5)^3+1/10*E(5)^4 ], [ 1/2, 1/10*E(5)-1/10*E(5)^2-1/10*E(5)^3+1/10*E(5)^4, -1/2, -3/10*E(5)+3/10*E(5)^2+3/10*E(5)^3-3/10*E(5)^4 ], [ 1/2, -1/10*E(5)+1/10*E(5)^2+1/10*E(5)^3-1/10*E(5)^4, -1/2, 3/10*E(5)-3/10*E(5)^2-3/10*E(5)^3+3/10*E(5)^4 ], [ 1/2, -3/10*E(5)+3/10*E(5)^2+3/10*E(5)^3-3/10*E(5)^4, 1/2, -1/10*E(5)+1/10*E(5)^2+1/10*E(5)^3-1/10*E(5)^4 ] ]
HaarTransform( n )
shortcut: HT
HaarTransform
returns the Haar transform on n points,
which must be a power of 2, <n> = 2k.
For a definition of the Haar transform see ER82, 10.10.
gap> HT(4); [ [ 1/4, 1/4, 1/4, 1/4 ], [ 1/4, 1/4, -1/4, -1/4 ], [ 1/4*E(8)-1/4*E(8)^3, -1/4*E(8)+1/4*E(8)^3, 0, 0 ], [ 0, 0, 1/4*E(8)-1/4*E(8)^3, -1/4*E(8)+1/4*E(8)^3 ] ]
InverseHaarTransform( n )
shortcut: InvHT
InverseHaarTransform
returns the inverse of the
Haar transform on n points,
which must be a power of 2, <n> = 2k.
The inverse is exactly n times the transpose of HT.
gap> InvHT(4); [ [ 1, 1, E(8)-E(8)^3, 0 ], [ 1, 1, -E(8)+E(8)^3, 0 ], [ 1, -1, 0, E(8)-E(8)^3 ], [ 1, -1, 0, -E(8)+E(8)^3 ] ]
74.145 RationalizedHaarTransform
RationalizedHaarTransform( n )
shortcut: RHT
RationalizedHaarTransform
returns the rationalized
Haar transform on n points,
which must be a power of 2, <n> = 2k.
For a definition of the rationalized Haar transform see ER82, 10.11.
gap> RHT(4); [ [ 1, 1, 1, 1 ], [ 1, 1, -1, -1 ], [ 1, -1, 0, 0 ], [ 0, 0, 1, -1 ] ]
74.146 InverseRationalizedHaarTransform
InverseRationalizedHaarTransform( n )
shortcut: InvRHT
InverseRationalizedHaarTransform
returns the inverse of the
rationalized Haar transform on n points,
which must be a power of 2, <n> = 2k.
gap> InvRHT(4); [ [ 1/4, 1/4, 1/2, 0 ], [ 1/4, 1/4, -1/2, 0 ], [ 1/4, -1/4, 0, 1/2 ], [ 1/4, -1/4, 0, -1/2 ] ]
The decomposition of a matrix M with symmetry is a striking application of constructive representation theory and was the original motivation to create the package \AREP. Here, decomposition means that M is decomposed into a product of highly structured sparse matrices. Applied to matrices corresponding to discrete signal transforms such a decomposition may represent a fast algorithm for the signal transform.
For the definition of symmetry see Symmetry of Matrices.
The idea of decomposing a matrix with symmetry is due to Minkwitz Min95, Min93 and was further developed by the authors of this package. See Egn97, chapter 1 or Pue98, chapter 3 for a thorough explanation of the method.
The following three functions correspond to the three types
of symmetry considered in this package (see Symmetry of Matrices).
The functions are implemented in the file "arep/lib/algogen.g"
.
74.148 MatrixDecompositionByPermPermSymmetry
MatrixDecompositionByPermPermSymmetry( M )
Let M be a matrix or an amat (see AMats).
MatrixDecompositionByPermPermSymmetry
returns a highly
structured amat of type "product"
with all factors being
sparse which represents the matrix M. The returned amat
can be viewed as a fast algorithm for the multiplication
with M.
The function uses the perm-perm symmetry (see PermPermSymmetry)
to decompose the matrix (see Matrix Decomposition) and
can hence be accelerated by setting UseLeon := true
as described in Symmetry of Matrices.
The following examples show that MatrixDecompositionByPermPermSymmetry
discovers automatically the method of Rader (see Rad68)
for a discrete Fourier transform of prime degree as well as
the well-known decomposition of circulant matrices.
gap> M := DFT(5);; gap> PrintArray(M); [ [ 1, 1, 1, 1, 1 ], [ 1, E(5), E(5)^2, E(5)^3, E(5)^4 ], [ 1, E(5)^2, E(5)^4, E(5), E(5)^3 ], [ 1, E(5)^3, E(5), E(5)^4, E(5)^2 ], [ 1, E(5)^4, E(5)^3, E(5)^2, E(5) ] ] gap> MatrixDecompositionByPermPermSymmetry(M); AMatPerm((4,5), 5) * DirectSumAMat( IdentityPermAMat(1), TensorProductAMat( DFTAMat(2), IdentityPermAMat(2) ) * DiagonalAMat([ 1, 1, 1, E(4) ]) * TensorProductAMat( IdentityPermAMat(2), DFTAMat(2) ) * AMatPerm((2,3), 4) ) * AMatPerm((1,4,2,5,3), 5) * DirectSumAMat( DiagonalAMat([ E(20)^4-E(20)^13-E(20)^16+E(20)^17, E(5)-E(5)^2-E(5)^3+E(5)^4, E(20)^4+E(20)^13-E(20)^16-E(20)^17 ]), AMatMat( [ [ 1, 4 ], [ 1, -1 ] ] ) ) * AMatPerm((1,3,5,2,4), 5) * DirectSumAMat( IdentityPermAMat(1), AMatPerm((2,3), 4) * TensorProductAMat( IdentityPermAMat(2), DiagonalAMat([ 1/2, 1/2 ]) * DFTAMat(2) ) * DiagonalAMat([ 1, 1, 1, -E(4) ]) * TensorProductAMat( DiagonalAMat([ 1/2, 1/2 ]) * DFTAMat(2), IdentityPermAMat(2) ) ) * AMatPerm((3,4,5), 5) gap> M := [[1, 2, 3], [3, 1, 2], [2, 3, 1]];; gap> PrintArray(M); [ [ 1, 2, 3 ], [ 3, 1, 2 ], [ 2, 3, 1 ] ] gap> MatrixDecompositionByPermPermSymmetry(M); DFTAMat(3) * AMatMon( Mon( (2,3), [ 2, 2/3*E(3)+1/3*E(3)^2, 1/3*E(3)+2/3*E(3)^2 ] ) ) * DFTAMat(3)
74.149 MatrixDecompositionByMonMonSymmetry
MatrixDecompositionByMonMonSymmetry( M )
Let M be a matrix or an amat (see AMats).
MatrixDecompositionByMonMonSymmetry
returns a highly
structured amat of type "product"
with all factors being
sparse which represents the matrix M. The returned amat
can be viewed as a fast algorithm for the multiplication
with M.
The function uses the mon-mon symmetry (see MonMonSymmetry)
to decompose the matrix (see Matrix Decomposition) and
can hence be accelerated by setting UseLeon := true
as described in Symmetry of Matrices.
The following example show that MatrixDecompositionByMonMonSymmetry
is able to find automatically a decomposition of the discrete
cosine transform of type IV (see DiscreteCosineTransformIV).
gap> M := DCT_IV(8);; gap> MatrixDecompositionByMonMonSymmetry(M); AMatMon( Mon( (3,4,7,6,8,5), [ E(4), E(16)^5, E(8)^3, -E(16)^7, 1, -E(16), E(8), -E(16)^3 ] ) ) * TensorProductAMat( DFTAMat(2), IdentityPermAMat(4) ) * DiagonalAMat([ 1, 1, 1, 1, 1, E(8), E(4), E(8)^3 ]) * TensorProductAMat( IdentityPermAMat(2), DFTAMat(2), IdentityPermAMat(2) ) * DiagonalAMat([ 1, 1, 1, E(4), 1, 1, 1, E(4) ]) * TensorProductAMat( IdentityPermAMat(4), DFTAMat(2) ) * DiagonalAMat([ -E(64), -E(64), E(64)^9, -E(64)^9, E(64)^23, -E(64)^23, E(64)^31, E(64)^31 ]) * TensorProductAMat( IdentityPermAMat(4), DiagonalAMat([ 1/2, 1/2 ]) * DFTAMat(2) ) * DiagonalAMat([ 1, 1, 1, -E(4), 1, 1, 1, -E(4) ]) * TensorProductAMat( IdentityPermAMat(2), DiagonalAMat([ 1/2, 1/2 ]) * DFTAMat(2), IdentityPermAMat(2) ) * DiagonalAMat([ 1, 1, 1, 1, 1, -E(8)^3, -E(4), -E(8) ]) * TensorProductAMat( DiagonalAMat([ 1/2, 1/2 ]) * DFTAMat(2), IdentityPermAMat(4) ) * AMatMon( Mon( (2,6,3,4,7,5,8), [ E(4), E(16)^5, -E(16)^7, E(8), E(8)^3, -E(16)^3, -E(16), 1 ] ) )
74.150 MatrixDecompositionByPermIrredSymmetry
MatrixDecompositionByPermIrredSymmetry( M )
MatrixDecompositionByPermIrredSymmetry( M, maxblocksize )
Let M be a matrix or an amat (see AMats).
MatrixDecompositionByPermIrredSymmetry
returns a highly
structured amat of type "product"
with all factors being
sparse which represents the matrix M. The returned amat
can be viewed as a fast algorithm for the multiplication
with M.
The function uses the perm-irred symmetry (see PermIrredSymmetry) to decompose the matrix (see Matrix Decomposition).
If maxblocksize is supplied only those perm-irred symmetries with all irreducibles having degree less than maxblocksize are considered. The default for maxblocksize is 2.
Note that the perm-irred symmetry is not symmetric. Hence it is possible that a matrix M decomposes but its transpose not.
The following examples show that
MatrixDecompositionByPermIrredSymmetry
discovers automatically the
Cooley-Tukey decomposition (see CT65) of a discrete Fourier
transform as well as a decomposition of the transposed
discrete cosine transform of type II (see DiscreteCosineTransform).
gap> M := DFT(4); [ [ 1, 1, 1, 1 ], [ 1, E(4), -1, -E(4) ], [ 1, -1, 1, -1 ], [ 1, -E(4), -1, E(4) ] ] gap> MatrixDecompositionByPermIrredSymmetry(M); TensorProductAMat( DFTAMat(2), IdentityPermAMat(2) ) * DiagonalAMat([ 1, 1, 1, E(4) ]) * TensorProductAMat( IdentityPermAMat(2), DFTAMat(2) ) * AMatPerm((2,3), 4) gap> M := TransposedMat(DCT(8));; gap> MatrixDecompositionByPermIrredSymmetry(M); AMatPerm((1,2,6,7,5,3,8), 8) * TensorProductAMat( IdentityPermAMat(2), AMatPerm((3,4), 4) * TensorProductAMat( IdentityPermAMat(2), DFTAMat(2) ) * AMatPerm((2,3), 4) * DirectSumAMat( DFTAMat(2), IdentityPermAMat(2) ) ) * AMatPerm((2,7,5,4,3)(6,8), 8) * DirectSumAMat( IdentityPermAMat(3), DirectSumAMat( IdentityPermAMat(1), AMatMat( [ [ -1/2*E(8)+1/2*E(8)^3, 1/2*E(8)-1/2*E(8)^3 ], [ 1/2*E(8)-1/2*E(8)^3, 1/2*E(8)-1/2*E(8)^3 ] ], "invertible" ) ), IdentityPermAMat(2) ) * DirectSumAMat( TensorProductAMat( DFTAMat(2), IdentityPermAMat(3) ), IdentityPermAMat(2) ) * AMatPerm((2,7,3,8,4), 8) * DirectSumAMat( DiagonalAMat([ 1/4*E(8)-1/4*E(8)^3, 1/4*E(8)-1/4*E(8)^3 ]), AMatMat( [ [ 1/4*E(16)-1/4*E(16)^7, 1/4*E(16)^3-1/4*E(16)^5 ], [ 1/4*E(16)^3-1/4*E(16)^5, -1/4*E(16)+1/4*E(16)^7 ] ] ), AMatMat( [ [ -1/4*E(32)+1/4*E(32)^15, -1/4*E(32)^7+1/4*E(32)^9 ], [ 1/4*E(32)^7-1/4*E(32)^9, -1/4*E(32)+1/4*E(32)^15 ] ] ), AMatMat( [ [ -1/4*E(32)^3+1/4*E(32)^13, -1/4*E(32)^5+1/4*E(32)^11 ], [ -1/4*E(32)^5+1/4*E(32)^11, 1/4*E(32)^3-1/4*E(32)^13 ] ] ) ) * AMatPerm((2,5)(4,7)(6,8), 8)
The next sections describe basic functions for the calculation with complex numbers which are represented as cyclotomics, e.g. computation of the complex conjugate or certain sine and cosine expressions.
The following functions are implemented in the file
"arep/lib/complex.g"
.
ImaginaryUnit( )
ImaginaryUnit
returns E(4)
.
gap> ImaginaryUnit(); E(4)
Re( z )
Re
returns the real part of the cyclotomic z.
gap> z := E(3) + E(4); E(12)^4-E(12)^7-E(12)^11 gap> Re(z); -1/2
Re( list )
Re
returns the list of the real parts of the cyclotomics
in list.
Im( z )
Im
returns the imaginary part of the cyclotomic z.
gap> z := E(3) + E(4); E(12)^4-E(12)^7-E(12)^11 gap> Im(z); -E(12)^4-1/2*E(12)^7-E(12)^8+1/2*E(12)^11
Im( list )
Im
returns the list of the imaginary parts of the cyclotomics
in list.
AbsSqr( z )
AbsSqr
returns the squared absolute value of the cyclotomic z.
gap> AbsSqr(z); -2*E(12)^4-E(12)^7-2*E(12)^8+E(12)^11
AbsSqr( list )
AbsSqr
returns the list of the squared absolute values
of the cyclotomics in list.
Sqrt( r )
Sqrt
returns the square root of the rational number r.
gap> Sqrt(1/3); 1/3*E(12)^7-1/3*E(12)^11
ExpIPi( r )
Let r be a rational number. ExpIPi
returns eπ i r.
gap> ExpIPi(1/5); -E(5)^3
CosPi( r )
Let r be a rational number. CosPi( r )
returns cos(π r).
gap> CosPi(1/5); -1/2*E(5)^2-1/2*E(5)^3
SinPi( r )
Let r be a rational number. SinPi( r )
returns sin(π r).
gap> SinPi(1/5); -1/2*E(20)^13+1/2*E(20)^17
TanPi( r )
Let r be a rational number. TanPi( r )
returns tan(π r).
gap> TanPi(1/5); E(20)-E(20)^9+E(20)^13-E(20)^17
74.161 Functions for Matrices and Permutations
The following sections describe basic functions for matrices and permutations, like forming the tensor product (Kronecker product) or direct sum and determination of the blockstructure of a matrix.
The following functions are implemented in the files
"arep/lib/permblk.g"
(kbs
, see kbs),
"arep/lib/summands.g"
(DirectSummandsPermutedMat
,
see DirectSummandsPermutedMat) and the file "arep/lib/tools.g"
(the other functions).
TensorProductMat( M1, ..., Mk )
TensorProductMat
returns the tensor product of the matrices
<M1>, ..., M_k.
gap> TensorProductMat( [[1]], [[1,2], [3,4]], [[5,6], [7,8]] ); [ [ 5, 6, 10, 12 ], [ 7, 8, 14, 16 ], [ 15, 18, 20, 24 ], [ 21, 24, 28, 32 ] ]
TensorProductMat( list )
TensorProductMat
returns the tensor product of the matrices
in list.
MatPerm( p, d )
MatPerm( p, d, char )
MatPerm
returns the permutation matrix of degree d
corresponding to the permutation p in characteristic char.
The default characteristic is 0. If d is less than the largest
moved point of p an error is signaled.
We use the following convention to create a permutation
gap> MatPerm( (1,2,3), 4 ); [ [ 0, 1, 0, 0 ], [ 0, 0, 1, 0 ], [ 1, 0, 0, 0 ], [ 0, 0, 0, 1 ] ]
PermMat( M )
PermMat
returns the permutation represented by the matrix M
and returns false otherwise. For the convention see MatPerm.
gap> PermMat( [[0,0,1], [1,0,0], [0,1,0]] ); (1,3,2)
PermutedMat( p1, M, p2 )
Let p1, p2 be permutations and M a matrix
with r rows and c columns.
PermutedMat
returns MatPerm( p1, r )
. M.MatPerm( p2, c )
(see MatPerm).
The largest moved point of p1 and p2 must
not exceed r resp. c otherwise an error is signaled.
gap> PermutedMat( (1,2), [[1,2,3], [4,5,6], [7,8,9]], (1,2,3) ); [ [ 6, 4, 5 ], [ 3, 1, 2 ], [ 9, 7, 8 ] ]
74.166 DirectSummandsPermutedMat
DirectSummandsPermutedMat( M )
DirectSummandsPermutedMat( M, match-blocks )
Let M be a matrix. DirectSummandsPermutedMat
returns the list
[ p1, [ M1, ..., Mk ], p2 ]
where
p1, p2 are permutations and Mi, i = 1,...,k, are
matrices with the property
<M> = PermutedMat( p_1, DiagonalMat( M_1, ..., M_k ),
p_2 )
(see PermutedMat, DiagonalMat).
If match-blocks is true
or not provided then the permutations
p1 and p2 are chosen such that equivalent Mi
are equal and occur next to each other.
If match-blocks is false
this is not done.
For an explanation of the algorithm see Egn97.
gap> M := [ [ 0, 0, 0, 2, 0, 1], [ 3, 1, 0, 0, 0, 0], > [ 0, 0, 1, 0, 2, 0], [ 1, 2, 0, 0, 0, 0], > [ 0, 0, 0, 1, 0, 3], [ 0, 0, 3, 0, 1, 0] ];; gap> PrintArray(M); [ [ 0, 0, 0, 2, 0, 1 ], [ 3, 1, 0, 0, 0, 0 ], [ 0, 0, 1, 0, 2, 0 ], [ 1, 2, 0, 0, 0, 0 ], [ 0, 0, 0, 1, 0, 3 ], [ 0, 0, 3, 0, 1, 0 ] ] gap> DirectSummandsPermutedMat(M); [ (2,4,3,5), [ [ [ 2, 1 ], [ 1, 3 ] ], [ [ 2, 1 ], [ 1, 3 ] ], [ [ 2, 1 ], [ 1, 3 ] ] ], (1,4)(2,6,3) ]
kbs( M )
Let M be a square matrix of degree n.
kbs
(konjugierte Blockstruktur = conjugated block structure)
returns the partition kbs(M) = {1,...,n}/R* where
R is the reflexive, symmetric, transitive closure of the relation
R defined by (i,j)∈ R⇔ M[i][j]\neq 0
.
For an investigation of the kbs of a matrix see Egn97.
gap> M := [[1,0,1,0], [0,2,0,3], [1,0,3,0], [0,4,0,1]]; [ [ 1, 0, 1, 0 ], [ 0, 2, 0, 3 ], [ 1, 0, 3, 0 ], [ 0, 4, 0, 1 ] ] gap> PrintArray(M); [ [ 1, 0, 1, 0 ], [ 0, 2, 0, 3 ], [ 1, 0, 3, 0 ], [ 0, 4, 0, 1 ] ] gap> kbs(M); [ [ 1, 3 ], [ 2, 4 ] ]
kbs( list )
kbs
returns the joined kbs of the matrices in list.
The matrices in list must have common size otherwise an
error is signaled.
74.168 DirectSumPerm
DirectSumPerm( list1, list2 )
Let list2 contain permutations and list1 be
of the same length and contain degrees equal or larger
than the corresponding largest moved points.
DirectSumPerm
returns the direct sum of the permutations
defined via the direct sum of the corresponding matrices.
gap> DirectSumPerm( [3, 3], [(1,2), (1,2,3)] ); (1,2)(4,5,6)
TensorProductPerm( list1, list2 )
Let list2 contain permutations and list1 be
of the same length and contain degrees equal or larger
than the corresponding largest moved points.
TensorProductPerm
returns the tensor product (Kronecker product)
of the permutations defined via the tensor product
of the corresponding matrices.
gap> TensorProductPerm( [3, 3], [(1,2), (1,2,3)] ); (1,5,3,4,2,6)(7,8,9)
gap3-jm