merge_to_overview

hybrid_learning.experimentation.ca_exp_eval.merge_to_overview(best_ious, all_stats, metric='set_iou', layers=None)

Merge the stats from all runs and from the best embeddings to an overview over the test values of metric. The resulting DataFrame is indexed by the layer_id, and columns are a multi-index of (concept_name, stats_name), with stats_name one of

• best_emb: the best embedding performance

• mean: the mean performance of the runs for that concept & layer

• std: the corresponding standard deviation

Parameters

• all_stats (DataFrame) – should an output of get_all_stats

• best_ious (DataFrame) – should an output of get_all_best_emb_stats

• metric (str) – the (column) name of the metric to use

• layers (Optional[List[str]]) – optionally restrict to given layers