| 摘要 |
<p>A method of multi-task learning for Bayesian matrix factorization includes receiving a plurality of datasets including a plurality of items rated by a plurality of users, each dataset representing a different task (401), receiving a parameter set (402), determining a posterior distribution for the parameter set given the datasets (403), wherein the posterior distribution is approximated by a factorization distribution (404), for determining a plurality of feature vectors, and outputting the feature vectors as a training model (405), wherein the trained model predicts a user's rating of an item.</p> |
