Collaborative filtering with maximum entropy

The collaborative filtering technique with maximum entropy, or maxnet, approach for generating online recommendations is discussed. Maxnet enables fast model querying and performs better than its competitors in terms of accuracy. Collaborative filtering methods work by assessing the similarities among users on the basis of their overlap in document requests, then recommending to a given user documents that like-minded users access previously. The model-based approach provides an attractive alternative to current recommenders in terms of the quality of its predictions.