ELBE

class mowl.models.ELBE(dataset, embed_dim=50, margin=0, reg_norm=1, learning_rate=0.001, batch_size=32768, model_filepath=None, device='cpu', neg_sampling_gcis=None)

Bases: EmbeddingELModel

Implementation based on [peng2020].

This model uses MSE loss to train the embeddings, where positive samples should have scores close to 0 and negative samples should have scores close to 1.

Methods Summary

compute_loss(pos_scores[, neg_scores])	Compute MSE loss for ELBE.
init_module()

Methods Documentation

compute_loss(pos_scores, neg_scores=None)

Compute MSE loss for ELBE.

Positive samples should have scores close to 0. Negative samples should have scores close to 1.

init_module()