No abstract

International audienceAlthough online recommendation systems such as recommendation of movies or music have been systematically studied in the past decade, location recommendation in Location Based Social Networks (LBSNs) is not well investigated yet. In LBSNs, users can check in and leave tips commenting on a venue. These two heterogeneous data sources both describe users' preference of venues. However, in current research work, only users' check-in behavior is considered in users' location preference model, users' tips on venues are seldom investigated yet. Moreover, while existing work mainly considers social influence in recommendation, we argue that considering venue similarity can further improve the recommendation performance. In this research, we ameliorate location recommendation by enhancing not only the user location preference model but also recommendation algorithm. First, we propose a hybrid user location preference model by combining the preference extracted from check-ins and text-based tips which are processed using sentiment analysis techniques. Second, we develop a location based social matrix factorization algorithm that takes both user social influence and venue similarity influence into account in location recommendation. Using two datasets extracted from the location based social networks Foursquare, experiment results demonstrate that the proposed hybrid preference model can better characterize user preference by maintaining the preference consistency, and the proposed algorithm outperforms the state-of-the-art methods

Deep Reinforcement Learning (DRL) has achieved impressive success in many applications. A key component of many DRL models is a neural network representing a Q function, to estimate the expected cumulative reward following a state-action pair. The Q function neural network contains a lot of implicit knowledge about the RL problems, but often remains unexamined and uninterpreted. To our knowledge, this work develops the first mimic learning framework for Q functions in DRL. We introduce Linear Model U-trees (LMUTs) to approximate neural network predictions. An LMUT is learned using a novel on-line algorithm that is well-suited for an active play setting, where the mimic learner observes an ongoing interaction between the neural net and the environment. Empirical evaluation shows that an LMUT mimics a Q function substantially better than five baseline methods. The transparent tree structure of an LMUT facilitates understanding the network's learned knowledge by analyzing feature influence, extracting rules, and highlighting the super-pixels in image inputs.

No abstract

No abstract