API Reference#

This is the class and function reference of GraKeL. In order for the user to understand how to use the package, we suggest he reads Documentation section.

grakel.graph: Graph class with its utility functions#

Base Class#

Graph([initialization_object, node_labels, ...])

The general graph class.

Utility Functions#

graph.is_adjacency(g[, transform])

Define if input is in a valid adjacency matrix format.

graph.is_edge_dictionary(g[, transform])

Define if input is in a valid edge dictionary format.

graph.laplacian(csgraph[, normed, ...])

Re-export of scipy.sparse.csgraph.laplacian.

graph.floyd_warshall(adjacency_matrix)

Calculate the Floyd Warshall, shortest path matrix.

User guide: See the Graph (class) section for further details.

grakel.graph_kernels: A kernel decorator#

Graph Kernel (decorator)#

grakel.GraphKernel([kernel, normalize, ...])

A generic wrapper for graph kernels.

User guide: See the GraphKernel (class) section for further details.

grakel.kernels: A collection of graph kernels#

Kernels#

Kernel([n_jobs, normalize, verbose])

A general class for graph kernels.

RandomWalk([n_jobs, normalize, verbose, ...])

The random walk kernel class.

RandomWalkLabeled([n_jobs, normalize, ...])

The labeled random walk kernel class.

PyramidMatch([n_jobs, normalize, verbose, ...])

Pyramid match kernel class.

NeighborhoodHash([n_jobs, normalize, ...])

Neighborhood hashing kernel as proposed in [HK09].

ShortestPath([n_jobs, normalize, verbose, ...])

The shortest path kernel class.

ShortestPathAttr([n_jobs, normalize, ...])

The shortest path kernel for attributes.

GraphletSampling([n_jobs, normalize, ...])

The graphlet sampling kernel.

SubgraphMatching([n_jobs, verbose, ...])

Calculate the subgraph matching kernel.

WeisfeilerLehman([n_jobs, verbose, ...])

Compute the Weisfeiler Lehman Kernel.

HadamardCode([n_jobs, verbose, normalize, ...])

The simple Hadamard code kernel, as proposed in [KI16].

NeighborhoodSubgraphPairwiseDistance([...])

The Neighborhood subgraph pairwise distance kernel.

LovaszTheta([n_jobs, normalize, verbose, ...])

Lovasz theta kernel as proposed in [JJDB14].

SvmTheta([n_jobs, normalize, verbose, ...])

Calculate the SVM theta kernel.

Propagation([n_jobs, verbose, normalize, ...])

The Propagation kernel for fully labeled graphs.

PropagationAttr([n_jobs, verbose, ...])

The Propagation kernel for fully attributed graphs.

OddSth([n_jobs, normalize, verbose, h])

ODD-Sth kernel as proposed in [DSMNS12].

MultiscaleLaplacian([n_jobs, normalize, ...])

Laplacian Graph Kernel as proposed in [KP16].

HadamardCode([n_jobs, verbose, normalize, ...])

The simple Hadamard code kernel, as proposed in [KI16].

VertexHistogram([n_jobs, normalize, ...])

Vertex Histogram kernel as found in [SB15].

EdgeHistogram([n_jobs, normalize, verbose, ...])

Edge Histogram kernel as found in [SB15].

GraphHopper([n_jobs, normalize, verbose, ...])

Graph Hopper Histogram kernel as found in [FKP+13].

CoreFramework([n_jobs, verbose, normalize, ...])

The core kernel framework, as proposed in [NMLV18].

WeisfeilerLehmanOptimalAssignment([n_jobs, ...])

Compute the Weisfeiler Lehman Optimal Assignment Kernel.

User guide: See the Kernels (between graphs) section for further details.

grakel.datasets: Datasets#

Fetch#

fetch_dataset(name[, verbose, data_home, ...])

Access a large collection of benchmark datasets from TU Dortmund [KKM+16].

get_dataset_info(dataset_name[, default])

Return the info concerning the existence of a certain dataset.

User guide: See the Dataset loading utilities section for further details.

grakel: Utils#

Use a kernel matrix as a transformer#

KMTransformer([K])

A Kernel Matrix Transformer.

Cross Validation#

cross_validate_Kfold_SVM(K, y[, n_iter, ...])

Cross Validate a list of precomputed kernels with an SVM.

Load from other file formats#

graph_from_networkx(X[, node_labels_tag, ...])

Transform an iterable of networkx objects to an iterable of Graphs.

graph_from_pandas(edge_df[, node_df, ...])

Produces a collection of Graph Objects from pandas dataframes.

graph_from_csv(edge_files[, node_files, ...])

Produces a collection of Graph Objects from a collection of csv files.

User guide: Usefull functions for applying to existing datasets, of other formats.