Retrieval metrics measure the quality of the embedding space given a
set query examples and a set of indexed examples. Informally it can be thought of as how well the space is clustered among other things.
-
class BNDCG: Binary normalized discounted cumulative gain. -
class MapAtK: Mean Average precision - mAP@K is computed as. -
class PrecisionAtK: Precision@K is computed as. -
class RecallAtK: The metric learning version of Recall@K. -
class RetrievalMetric: Abstract base class for computing retrieval metrics.