18.5.15 Example: Computing Edge Embeddings on the Movielens Dataset

This section describes the usage of UnsupervisedEdgeWise in PGX using the Movielens graph as an example.

This data set consists of 100,000 ratings (1-5) from 943 users on 1682 movies, with simple demographic information for the users (age, gender, occupation) and movies (year, aggravating, genre). Users and movies are vertices, while ratings of users to movies are edges with a rating feature.

The following example predicts the ratings using the UnsupervisedEdgeWise model. You first build the model and fit it on the trainGraph.

• JShell
• Java
• Python

opg4j> var convLayer = analyst.graphWiseConvLayerConfigBuilder().
        setNumSampledNeighbors(10).
        build()

opg4j> var model = analyst.unsupervisedEdgeWiseModelBuilder().
        setVertexInputPropertyNames("movie_year", "avg_rating", "movie_genres", // Movies features
            "user_occupation_label", "user_gender", "raw_user_age"). // Users features
        setEdgeInputPropertyNames("user_rating").
        setConvLayerConfigs(convLayer).
        setNumEpochs(10).
        setEmbeddingDim(32).
        setLearningRate(0.003).
        setStandardize(true).
        setNormalize(false). //recommended
        setSeed(0).
        build()
opg4j> model.fit(trainGraph)

GraphWiseConvLayerConfig convLayer = analyst.graphWiseConvLayerConfigBuilder()
        .setNumSampledNeighbors(10)
        .build();

UnsupervisedEdgeWiseModel model = analyst.unsupervisedEdgeWiseModelBuilder()
        .setVertexInputPropertyNames("movie_year", "avg_rating", "movie_genres", // Movies features
            "user_occupation_label", "user_gender", "raw_user_age") // Users features
        .setEdgeInputPropertyNames("user_rating")
        .setConvLayerConfigs(convLayer)
        .setNumEpochs(10)
        .setEmbeddingDim(32)
        .setLearningRate(0.003)
        .setStandardize(true)
        .setNormalize(false) //recommended
        .setSeed(0)
        .build();

model.fit(trainGraph);

conv_layer_config = dict(num_sampled_neighbors=10)

conv_layer = analyst.graphwise_conv_layer_config(**conv_layer_config)

params = dict(conv_layer_config=[conv_layer],
              vertex_input_property_names=["movie_year", "avg_rating", "movie_genres",
                "user_occupation_label", "user_gender", "raw_user_age"],
              edge_input_property_names=["user_rating"],
              num_epochs=10,
              embedding_dim=32,
              learning_rate=0.003,
              normalize=False, #recommended
              seed=0)

model = analyst.unsupervised_edgewise_builder(**params)

model.fit(train_graph)

Since EdgeWise is inductive, you can infer the ratings for unseen edges:

• JShell
• Java
• Python

opg4j> var embeddings = model.inferEmbeddings(fullGraph, testEdges)
opg4j> embeddings.head().print()

PgxFrame embeddings = model.inferEmbeddings(fullGraph,testEdges);
embeddings.head().print();

embeddings = model.infer_embeddings(full_graph, test_edges)
embeddings.print()