Neural choice by elimination via highway networks

We introduce Neural Choice by Elimination, a new framework that integrates deep neural networks into probabilistic sequential choice models for learning to rank. Given a set of items to chose from, the elimination strategy starts with the whole item set and iteratively eliminates the least worthy item in the remaining subset. We prove that the choice by elimination is equivalent to marginalizing out the random Gompertz latent utilities. Coupled with the choice model is the recently introduced Neural Highway Networks for approximating arbitrarily complex rank functions. We evaluate the proposed framework on a large-scale public dataset with over 425K items, drawn from the Yahoo! learning to rank challenge. It is demonstrated that the proposed method is competitive against state-of-the-art learning to rank methods.

Exploring the effectiveness of true abnormal data elimination in context-aware web services recommendation

Recent years have witnessed a growing interest in context-aware recommender system (CARS), which explores the impact of context factors on personalized Web services recommendation. Basically, the general idea of CARS methods is to mine historical service invocation records through the process of context-aware similarity computation. It is observed that traditional similarity mining process would very likely generate relatively big deviations of QoS values, due to the dynamic change of contexts. As a consequence, including a considerable amount of deviated QoS values in the similarity calculation would probably result in a poor accuracy for predicting unknown QoS values. In allusion to this problem, this paper first distinguishes two definitions of Abnormal Data and True Abnormal Data, the latter of which should be eliminated. Second, we propose a novel CASR-TADE method by incorporating the effectiveness of True Abnormal Data Elimination into context-aware Web services recommendation. Finally, the experimental evaluations on a real-world Web services dataset show that the proposed CASR-TADE method significantly outperforms other existing approaches.