Models¶

HyperTorch provides ready-to-use built-in models inspired by the existing literature.

At a high level: - hypertorch.hlp.* contains ready-to-train hyperlink prediction (HLP) modules. - hypertorch.nc.* contains ready-to-train node classification (NC) modules. - hypertorch.models.* contains actual models like Node2Vec, GCN, etc. - hypertorch.nn.* contains layers, enrichers, aggregators, and losses.

Built-in hyperlink prediction modules¶

Supported models include:

• CommonNeighbors (non-trainable baseline).
• GCN.
• HGNN.
• HGNNP.
• HNHN.
• HyperGCN.
• MLP.
• NHP.
• Node2VecGCN.
• Node2VecSLP.
• VilLain.

Built-in node classification modules¶

Supported models include:

• CommonNeighbors (non-trainable baseline).
• GCN.
• HGNN.
• HGNNP.
• HNHN.
• HyperGCN.
• MLP.

Minimal hyperlink prediction example: NHP¶

from hypertorch.hlp import NHPHlpModule

model = NHPHlpModule(
    encoder_config={
        "in_channels": num_features,
        "hidden_channels": 512,
        "aggregation": "maxmin",
    },
    lr=0.001,
    weight_decay=5e-4,
    metrics=metrics,
)

Minimal node classification example: HyperGCN¶

from hypertorch.nc import HyperGCNNcModule

model = HyperGCNNcModule(
    classifier_config={
        "in_channels": 32,
        "hidden_channels": 16,
        "out_channels": 3,
        "drop_rate": 0.3,
        "use_mediator": False,
    },
    lr=0.01,
    weight_decay=5e-4,
)

Next steps¶

• Training loop (callbacks, devices, etc.): Training.
• Comparing multiple models consistently: Benchmarking.
• Outputs and logging: Loggers.
• Visualizing runs: TensorBoard.