Metrics

We have implemented a user-friendly class for evaluation of disentanglement with respect to known discrete targets.

metrics.DiscreteDisentanglementBenchmark

Benchmark for evaluating discrete disentanglement in latent representations.

The following functions represent the similarity functions used in benchmarking:

metrics.nn_alignment_score	Compute nearest neighbor alignment scores for all continuous variables.
metrics.local_mutual_info_score	Compute local mutual information scores for all variables and categories.
metrics.spearman_correlataion_score	Compute Spearman correlation scores between continuous variables and categories.

The following functions represent the aggregation functions used in benchmarking:

metrics.most_similar_averaging_score	Compute the most similar averaging score for disentanglement evaluation.
metrics.latent_matching_score	Compute the latent matching score for disentanglement evaluation.
metrics.most_similar_gap_score	Compute the most similar gap score for disentanglement evaluation.