View source on GitHub |
Ranking policy.
Classes
class CosinePenalizedPlackettLuce
: A distribution that samples items based on scores and cosine similarity.
class DescendingScoreRankingPolicy
: A policy that is deterministically ranks elements based on their scores.
class DescendingScoreSampler
: Base neural network module class.
class NoPenaltyPlackettLuce
: Identical to PlackettLuce, with input signature modified to our needs.
class NoPenaltyRankingPolicy
: A class implementing ranking policies in TF Agents.
class PenalizeCosineDistanceRankingPolicy
: A Ranking policy that penalizes scores based on cosine distance.
class PenalizedPlackettLuce
: A distribution that samples permutations and penalizes item scores.
class RankingPolicy
: A class implementing ranking policies in TF Agents.