Data retrieval using reinforced co-learning for semi-supervised ranking
AssigneeSnap Inc.
InventorsShibi He, Yanen Li, Ning Xu
Life of this patent
Estimating the term…
Abstract
A computer-implement method comprises: training a classifier with labeled data from a dataset; classifying, by the trained classifier, unlabeled data from the dataset; providing, by the classifier to a policy gradient, a reward signal for each data/query pair; transferring, by the classifier to a ranker, learning; training, by the policy gradient, the ranker; ranking data from the dataset based on a query; and retrieving data from the ranked data in response to the query.
Claims20
A computer-implemented method, comprising:
- classifying, by a trained classifier, unlabeled data from a dataset, wherein the classifier includes a neural network;
- providing iteratively, by the classifier to a policy gradient function, a reward signal for data/query pairs, wherein the reward signal is a combination of a normalized discounted cumulative gain and a discriminative score output by the classifier;
- transferring, by the classifier to a ranker, learning from the classifying, wherein the ranker includes a neural network;
- training, by the policy gradient function, the ranker;
- and
- ranking, by the trained ranker, data from the dataset based on a query.
A system, comprising:
- one or more processors of a machine;
- a memory storing instruction that, when executed by the one or more processors, cause the machine to perform operations comprising:
- classifying, by a trained classifier, unlabeled data from a dataset, wherein the classifier includes a neural network with feature sharing;
- providing iteratively, by the classifier to a policy gradient function, a reward signal for data/query pairs, wherein the reward signal is a combination of a normalized discounted cumulative gain and a discriminative score output by the classifier;
- transferring, by the classifier to a ranker, learning from the classifying;
- training, by the policy gradient function, the ranker;
- and
- ranking, by the trained ranker, data from the dataset based on a query.
A machine-readable storage device embodying instructions that, when executed by a machine, cause the machine to perform operations comprising:
- classifying, by a trained classifier, unlabeled data from a dataset;
- providing iteratively, by the classifier to a policy gradient function, a reward signal for data/query pairs, wherein the reward signal is a combination of a normalized discounted cumulative gain and a discriminative score output by the classifier;
- transferring, by the classifier to a ranker, learning from the classifying;
- training, by the policy gradient function, the ranker;
- ranking, by the trained ranker, data from the dataset based on a query;
- and
- retrieving ranked data from the dataset.
Dependencies are inferred from the claim text.