Table 3 Intervals and values used in the hyperparameter study we carried out to optimize the TransE, ComplEx, and RotatE baselines
From: BioBLP: a modular framework for learning on multimodal biomedical knowledge graphs
Hyperparameter | Values |
|---|---|
Learning rate | Log-uniformly sampled from \([1\times 10^{-3}, 1.0]\) |
Regularization weight | Log-uniformly sampled from \([1\times 10^{-6}, 1\times 10^{-3}]\) |
Batch size | \(\lbrace 128, 256, 512, 1024\rbrace\) |
Loss function | Margin ranking, binary cross-entropy, cross-entropy |