KGEmbeddingBenchmark

Framework for embedding whole Knowledge Graphs and compare various techniques.

Knowledge Graph Comparison

This frameworks allows to generate and compare whole Knowledge Graph embeddings based on six graph embedding techniques:

• Graph2Vec
• GE-FSG
• Deep Graph Kernels
• weisfeiler lehman kernel
• SimGNN
• Graph Matching Networks

Usage:

 graphs = ...
 # Graph2Vec
 embedder = KGEmbedder(graphs,'temp_dir/',clear_dir=True)
 # GE-FSG
 embedder = GEFSGEmbedder(graphs,'temp_dir/',min_sup=2,clear_dir=True)
 # DGK
 embedder = DeepGraphEmbedder(graphs)
 # WL
 embedder = WLEmbedder(graphs,'temp_dir/')
 # SimGNN
 # requires train and test set of graphs with their GED
 embedder = SimGNNEmbedder(graphs_train,graphs_test,output_dir_train,output_dir_test,relable=True,clear_dir=True)
 # GMN
 embedder = GMNEmbedder(graphs)
 
 embeddings = embedder.embed()