[Up] [Previous] [Next] [Index]

# 8 Automorphism groups and isomorphism testing for graphs

### Sections

GRAPE includes B. D. McKay's nauty (Version 2.2, final patched version) package for calculating automorphism groups of graphs and for testing graph isomorphism (see Nau90,MP14). As described in Section Installing the GRAPE Package, a user may instead use their own copy of dreadnaut or dreadnautB from a later version of nauty, or may use T. Juntilla's and P. Kaski's bliss package JK07 instead of nauty. Many functions described in this chapter make use of nauty or bliss.

## 8.1 Graphs with colour-classes

For each of the functions described in this chapter, each graph parameter may be replaced by a graph with colour-classes, which is a record having (at least) the components graph (which should be a graph in GRAPE format), and colourClasses, which should be an ordered partition of the vertices of the graph, and so define colour-classes for the vertices. This ordered partition should be given as a list of (pairwise-disjoint non-empty) sets partitioning the vertex-set. When these functions are called with graphs with colour-classes, then it is understood that an automorphism of a graph with colour-classes is an automorphism of the graph which additionally preserves the list of colour-classes (classwise), and an isomorphism from one graph with colour-classes to a second is a graph isomorphism from the first graph to the second which additionally maps the first list of colour-classes to the second (classwise). The record for a graph with colour-classes may also optionally contain the additional components autGroup and/or canonicalLabelling, and these are handled in an analogous way to those for a graph (such as when using the parameter firstunbindcanon). Note that we do not require that adjacent vertices be in different colour-classes.

## 8.2 AutGroupGraph

• AutGroupGraph( gamma )
• AutGroupGraph( gamma, colourclasses )

The first version of this function returns the automorphism group of the graph (or graph with colour-classes) gamma, using nauty or bliss (this can also be accomplished by typing AutomorphismGroup(gamma)). The automorphism group Aut(gamma) of a graph gamma is the group consisting of the permutations of the vertices of gamma which preserve the edge-set of gamma. The automorphism group of a graph with colour-classes is the subgroup of the automorphism group of the graph which preserves the colour-classes (classwise).

The second version of this function is maintained only for backward compatibility. For this version gamma must be a graph, colourclasses is an ordered partition of the vertices of gamma, and the subgroup of Aut(gamma) preserving this ordered partition is returned. The ordered partition should be given as a list of (pairwise-disjoint non-empty) sets partitioning the vertices of gamma, although for backward compatibility and only in this situation, the last set in the ordered partition need not be included explicitly.

gap> gamma := JohnsonGraph(4,2);
rec( adjacencies := [ [ 2, 3, 4, 5 ] ],
group := Group([ (1,4,6,3)(2,5), (2,4)(3,5) ]), isGraph := true,
isSimple := true,
names := [ [ 1, 2 ], [ 1, 3 ], [ 1, 4 ], [ 2, 3 ], [ 2, 4 ], [ 3, 4 ] ],
order := 6, representatives := [ 1 ],
schreierVector := [ -1, 2, 1, 1, 1, 1 ] )
gap> Size(AutGroupGraph(gamma));
48
gap> AutGroupGraph( rec(graph:=gamma,colourClasses:=[[1,2,3],[4,5,6]]) );
Group([ (2,3)(4,5), (1,2)(5,6) ])
gap> Size(AutomorphismGroup( rec(graph:=gamma,colourClasses:=[[1,6],[2,3,4,5]]) ));
16

## 8.3 GraphIsomorphism

• GraphIsomorphism( gamma1, gamma2 )
• GraphIsomorphism( gamma1, gamma2, firstunbindcanon )

Let gamma1 and gamma2 both be graphs or both be graphs with colour-classes. Then this function makes use of nauty or bliss to (try to) determine an isomorphism from gamma1 to gamma2. If gamma1 and gamma2 are isomorphic, then this function returns an isomorphism from gamma1 to gamma2. This isomorphism will be a permutation of the vertices of gamma1 which maps the edge-set of gamma1 onto that of gamma2, and if gamma1 and gamma2 are graphs with colour-classes, this isomorphism will also map the colour-class list of gamma1 to that of gamma2 (classwise). If gamma1 and gamma2 are not isomorphic then this function returns fail.

The optional boolean parameter firstunbindcanon determines whether or not the canonicalLabelling components of both gamma1 and gamma2 are first unbound before proceeding. If firstunbindcanon is true (the default, safe and possibly slower option) then these components are first unbound. If firstunbindcanon is false, then any existing canonicalLabelling components are used. However, since canonical labellings can depend on whether nauty or bliss is used, the version of nauty or bliss used, the version of GRAPE, parameter settings of nauty or bliss, and possibly even the compiler and computer used, you must be sure that if firstunbindcanon=false then the canonicalLabelling component(s) which may already exist for gamma1 or gamma2 were created in exactly the same environment in which you are presently computing.

Please also note that a canonical labelling for a GRAPE graph is the inverse of what a canononical labelling for a graph is usually defined as (such as in bliss), in that in GRAPE, the image of a graph under the inverse of its canonical labelling is the calculated canonical version of that graph.

gap> gamma := JohnsonGraph(5,3);
rec( adjacencies := [ [ 2, 3, 4, 5, 7, 8 ] ],
group := Group([ (1,7,10,6,3)(2,8,4,9,5), (4,7)(5,8)(6,9) ]),
isGraph := true, isSimple := true,
names := [ [ 1, 2, 3 ], [ 1, 2, 4 ], [ 1, 2, 5 ], [ 1, 3, 4 ], [ 1, 3, 5 ],
[ 1, 4, 5 ], [ 2, 3, 4 ], [ 2, 3, 5 ], [ 2, 4, 5 ], [ 3, 4, 5 ] ],
order := 10, representatives := [ 1 ],
schreierVector := [ -1, 1, 1, 2, 1, 1, 1, 2, 1, 1 ] )
gap> delta := JohnsonGraph(5,2);
rec( adjacencies := [ [ 2, 3, 4, 5, 6, 7 ] ],
group := Group([ (1,5,8,10,4)(2,6,9,3,7), (2,5)(3,6)(4,7) ]),
isGraph := true, isSimple := true,
names := [ [ 1, 2 ], [ 1, 3 ], [ 1, 4 ], [ 1, 5 ], [ 2, 3 ], [ 2, 4 ],
[ 2, 5 ], [ 3, 4 ], [ 3, 5 ], [ 4, 5 ] ], order := 10,
representatives := [ 1 ], schreierVector := [ -1, 2, 2, 1, 1, 1, 2, 1, 1, 1
] )
gap> GraphIsomorphism( gamma, delta );
(3,5,6,8,7,4)
gap> GraphIsomorphism(
>       rec(graph:=gamma, colourClasses:=[[7],[1,2,3,4,5,6,8,9,10]]),
>       rec(graph:=delta, colourClasses:=[[10],[1..9]]) );
(1,3)(2,6,5)(4,8)(7,10,9)
gap> GraphIsomorphism(
>       rec(graph:=gamma, colourClasses:=[[1],[6],[2,3,4,5,7,8,9,10]]),
>       rec(graph:=delta, colourClasses:=[[1],[6],[2,3,4,5,7,8,9,10]]) );
fail

## 8.4 IsIsomorphicGraph

• IsIsomorphicGraph( gamma1, gamma2 )
• IsIsomorphicGraph( gamma1, gamma2, firstunbindcanon )

Let gamma1 and gamma2 both be graphs or both be graphs with colour-classes. Then this boolean function makes use of the nauty or bliss package to test whether gamma1 and gamma2 are isomorphic (as graphs or as graphs with colour-classes, respectively). The value true is returned if and only if the graphs (or graphs with colour-classes) are isomorphic.

The optional boolean parameter firstunbindcanon determines whether or not the canonicalLabelling components of both gamma1 and gamma2 are first unbound before proceeding. If firstunbindcanon is true (the default, safe and possibly slower option) then these components are first unbound. If firstunbindcanon is false, then any existing canonicalLabelling components are used. However, since canonical labellings can depend on whether nauty or bliss is used, the version of nauty or bliss used, the version of GRAPE, parameter settings of nauty or bliss, and possibly even the compiler and computer used, you must be sure that if firstunbindcanon=false then the canonicalLabelling component(s) which may already exist for gamma1 or gamma2 were created in exactly the same environment in which you are presently computing.

See also GraphIsomorphism. For pairwise isomorphism testing of three or more graphs (or graphs with colour-classes), see GraphIsomorphismClassRepresentatives.

Please also note that a canonical labelling for a GRAPE graph is the inverse of what a canononical labelling for a graph is usually defined as (such as in bliss), in that in GRAPE, the image of a graph under the inverse of its canonical labelling is the calculated canonical version of that graph.

gap> gamma := JohnsonGraph(5,3);
rec( adjacencies := [ [ 2, 3, 4, 5, 7, 8 ] ],
group := Group([ (1,7,10,6,3)(2,8,4,9,5), (4,7)(5,8)(6,9) ]),
isGraph := true, isSimple := true,
names := [ [ 1, 2, 3 ], [ 1, 2, 4 ], [ 1, 2, 5 ], [ 1, 3, 4 ], [ 1, 3, 5 ],
[ 1, 4, 5 ], [ 2, 3, 4 ], [ 2, 3, 5 ], [ 2, 4, 5 ], [ 3, 4, 5 ] ],
order := 10, representatives := [ 1 ],
schreierVector := [ -1, 1, 1, 2, 1, 1, 1, 2, 1, 1 ] )
gap> delta := JohnsonGraph(5,2);
rec( adjacencies := [ [ 2, 3, 4, 5, 6, 7 ] ],
group := Group([ (1,5,8,10,4)(2,6,9,3,7), (2,5)(3,6)(4,7) ]),
isGraph := true, isSimple := true,
names := [ [ 1, 2 ], [ 1, 3 ], [ 1, 4 ], [ 1, 5 ], [ 2, 3 ], [ 2, 4 ],
[ 2, 5 ], [ 3, 4 ], [ 3, 5 ], [ 4, 5 ] ], order := 10,
representatives := [ 1 ], schreierVector := [ -1, 2, 2, 1, 1, 1, 2, 1, 1, 1
] )
gap> IsIsomorphicGraph( gamma, delta );
true
gap> IsIsomorphicGraph(
>       rec(graph:=gamma, colourClasses:=[[7],[1,2,3,4,5,6,8,9,10]]),
>       rec(graph:=delta, colourClasses:=[[10],[1..9]]) );
true
gap> IsIsomorphicGraph(
>       rec(graph:=gamma, colourClasses:=[[1],[6],[2,3,4,5,7,8,9,10]]),
>       rec(graph:=delta, colourClasses:=[[1],[6],[2,3,4,5,7,8,9,10]]) );
false

## 8.5 GraphIsomorphismClassRepresentatives

• GraphIsomorphismClassRepresentatives( L )
• GraphIsomorphismClassRepresentatives( L, firstunbindcanon )

Given a list L of graphs, or of graphs with colour-classes, this function uses nauty or bliss to return a list consisting of pairwise non-isomorphic elements of L, representing all the isomorphism classes of elements of L.

The optional boolean parameter firstunbindcanon determines whether or not the canonicalLabelling components of all elements of L are first unbound before proceeding. If firstunbindcanon is true (the default, safe and possibly slower option) then these components are first unbound. If firstunbindcanon is false, then any existing canonicalLabelling components of elements of L are used. However, since canonical labellings can depend on whether nauty or bliss is used, the version of nauty or bliss used, the version of GRAPE, parameter settings of nauty or bliss, and possibly even the compiler and computer used, you must be sure that if firstunbindcanon=false then any canonicalLabelling component(s) which may already exist for elements of L were created in exactly the same environment in which you are presently computing.

It is assumed that the computing environment is constant throughout the execution of this function.

Please also note that a canonical labelling for a GRAPE graph is the inverse of what a canononical labelling for a graph is usually defined as (such as in bliss), in that in GRAPE, the image of a graph under the inverse of its canonical labelling is the calculated canonical version of that graph.

gap> A:=JohnsonGraph(5,3);
rec( adjacencies := [ [ 2, 3, 4, 5, 7, 8 ] ],
group := Group([ (1,7,10,6,3)(2,8,4,9,5), (4,7)(5,8)(6,9) ]),
isGraph := true, isSimple := true,
names := [ [ 1, 2, 3 ], [ 1, 2, 4 ], [ 1, 2, 5 ], [ 1, 3, 4 ], [ 1, 3, 5 ],
[ 1, 4, 5 ], [ 2, 3, 4 ], [ 2, 3, 5 ], [ 2, 4, 5 ], [ 3, 4, 5 ] ],
order := 10, representatives := [ 1 ],
schreierVector := [ -1, 1, 1, 2, 1, 1, 1, 2, 1, 1 ] )
gap> B:=JohnsonGraph(5,2);
rec( adjacencies := [ [ 2, 3, 4, 5, 6, 7 ] ],
group := Group([ (1,5,8,10,4)(2,6,9,3,7), (2,5)(3,6)(4,7) ]),
isGraph := true, isSimple := true,
names := [ [ 1, 2 ], [ 1, 3 ], [ 1, 4 ], [ 1, 5 ], [ 2, 3 ], [ 2, 4 ],
[ 2, 5 ], [ 3, 4 ], [ 3, 5 ], [ 4, 5 ] ], order := 10,
representatives := [ 1 ], schreierVector := [ -1, 2, 2, 1, 1, 1, 2, 1, 1, 1
] )
gap> R:=GraphIsomorphismClassRepresentatives([A,B,ComplementGraph(A)]);;
gap> Length(R);
2
gap> List(R,VertexDegrees);
[ [ 6 ], [ 3 ] ]
gap> R:=GraphIsomorphismClassRepresentatives(
>    [ rec(graph:=gamma, colourClasses:=[[1],[6],[2,3,4,5,7,8,9,10]]),
>      rec(graph:=delta, colourClasses:=[[1],[6],[2,3,4,5,7,8,9,10]]),
>      rec(graph:=ComplementGraph(gamma), colourClasses:=[[1],[6],[2,3,4,5,7,8,9,10]]) ] );;
gap> Length(R);
3

[Up] [Previous] [Next] [Index]

grape manual
March 2021