Reinforcement Routing on Proximity Graph for Efficient Recommendation

We focus on Maximum Inner Product Search (MIPS), which is an essential problem in many machine learning communities. Given a query, MIPS finds the most similar items with maximum inner products. Methods for Nearest Neighbor (NNS) usually defined metric space do not exhibit satisfactory performance since product non-metric function. However, products good properties compared functions, such as avoiding vanishing and exploding gradients. As result, widely used recommendation systems, makes efficient key speeding up systems. Graph-based methods NNS show superiorities other class methods. Each data point of database mapped to node proximity graph. neighbor search can be converted route graph find nearest query. This technique solve problem. Instead searching we item query In this article, propose reinforcement model train agent automatically if lack ground truths training queries. If know some queries, our also utilize these by imitation improve agent’s searchability. By experiments, see that proposed mode combines shows over state-of-the-art