CayleyPy API Reference

Core classes and functions

cayleypy.CayleyGraphDef(generators_type, ...)	Mathematical definition of a CayleyGraph.
cayleypy.CayleyGraph(definition, *[, ...])	Represents a Schreier coset graph for some group.
cayleypy.CayleyPath(start_state, edges, graph)	Path in a Cayley graph.
cayleypy.find_path(graph, start_state, **kwargs)	Finds path from start_state to central state.

Graphs library

cayleypy.PermutationGroups()	Pre-defined Cayley graphs for permutation groups (S_n).
cayleypy.MatrixGroups()	Pre-defined Cayley graphs for matrix groups.
cayleypy.Puzzles()	Definitions of graphs describing various puzzles.
cayleypy.GapPuzzles()	Library of puzzles defined in GAP format.
cayleypy.create_graph(*[, ...])	Creates CayleyGraph.
cayleypy.prepare_graph(name[, n])	Returns pre-defined CayleyGraphDef by codename and additional kwargs.

Beam search and ML

cayleypy.Predictor(graph, models_or_heuristics)	Estimates distance from central state to given states.
cayleypy.algo.BeamSearchAlgorithm(graph)	Beam search algorithm for finding paths in Cayley graphs.
cayleypy.algo.BeamSearchResult(path_found, ...)	Result of running beam search on a Cayley graph.
cayleypy.algo.RandomWalksGenerator(graph)	Generator for random walks on Cayley graphs.
cayleypy.models.ModelConfig(model_type, ...)	Configuration used to describe ML model.
cayleypy.models.GroupTokenizer(groups)	Converts states to sequences of tokens, one token per group of consecutive elements of a state.
cayleypy.models.MlpModel(config)	Multi-layer perceptron model.
cayleypy.models.ResMlpModel(config)	Multi-layer perceptron with residual (skip) connections.
cayleypy.models.graph_hash(graph_def)	Computes hash of the mathematical definition of a graph.
cayleypy.models.save_checkpoint(path, model, ...)	Saves weights of a model together with the config describing this model.
cayleypy.models.load_checkpoint(path[, ...])	Loads model from a checkpoint written by save_checkpoint().

Training

cayleypy.train.Loss()	Base class for losses used to train distance-estimating models.
cayleypy.train.MseLoss()	Mean squared error loss.
cayleypy.train.PinballLoss([tau])	Pinball (quantile) loss, which penalizes underestimation and overestimation differently.
cayleypy.train.make_loss(name[, tau])	Creates loss by name.

BFS algorithm and its variations

cayleypy.algo.BfsAlgorithm()	Basic version of the bread-first search (BFS) algorithm.
cayleypy.algo.BfsDistributed()	Multi-GPU breadth-first search implementation.
cayleypy.algo.BfsResult(bfs_completed, ...)	Result of running breadth-first search on a Schreier coset graph.
cayleypy.algo.bfs_bitmask(graph[, max_diameter])	Version of BFS storing all vertices explicitly as bitmasks, using 3 bits of memory per state.
cayleypy.algo.bfs_numpy(graph[, max_diameter])	Simple version of BFS (from destination_state) using numpy, optimized for memory usage.
cayleypy.algo.InteractiveBfs(graph, start_states)	Interactive breadth-first search that computes layers one by one.
cayleypy.algo.MeetInTheMiddle()	Meet-in-the middle (MITM) algorithm for path finding.