Learn More
Convolutional neural networks (CNNs) are inherently limited to model geometric transformations due to the fixed geometric structures in its building modules. In this work, we introduce two new modules to enhance the transformation modeling capacity of CNNs, namely, deformable convolution and deformable RoI pooling. Both are based on the idea of augmenting(More)
Collaborative Filtering with Implicit Feedbacks (e.g., browsing or clicking records), named as CF-IF, is demonstrated to be an effective way in recommender systems. Existing works of CF-IF can be mainly classified into two categories, i.e., point-wise regression based and pairwise ranking based, where the latter one relaxes assumption and usually obtains(More)
Deep convolutional neutral networks have achieved great success on image recognition tasks. Yet, it is nontrivial to transfer the state-of-the-art image recognition networks to videos as per-frame evaluation is too slow and unaffordable. We present deep feature flow, a fast and accurate framework for video recognition. It runs the expensive convolutional(More)
Point-of-interest (POI) recommendation has become more and more important, since it could discover user behavior pattern and find interesting venues for them. To address this problem, we propose a rank-based method, PGRank, which integrates user geographical preference and latent preference into Bayesian personalized ranking framework. The experimental(More)
  • 1