Learning over No-Preferred and Preferred Sequence of Items for Robust Recommendation

Abstract

In this paper, we propose a theoretically supported sequential strategy for training a large-scale Recommender System (RS) over implicit feedback, mainly in the form of clicks. The proposed approach consists in minimizing pairwise ranking loss over blocks of consecutive items constituted by a sequence of non-clicked items followed by a clicked one for each user. We present two variants of this strategy where model parameters are updated using either the momentum method or a gradient-based approach. To prevent updating the parameters for an abnormally high number of clicks over some targeted items (mainly due to bots), we introduce an upper and a lower threshold on the number of updates for each user. These thresholds are estimated over the distribution of the number of blocks in the training set. They affect the decision of RS by shifting the distribution of items that are shown to the users. Furthermore, we provide a convergence analysis of both algorithms and demonstrate their practical efficiency over six large-scale collections with respect to various ranking measures and computational time.

Large-Scale Sequential Learning for Recommender and Engineering Systems

“…It is important to note that the proposed algorithm was theoretically justified. First we proved its convergence for the case of a convex loss function, and then we extend the theory of convergence on the general case [2].…”

“…Chapter 5, is based on two papers published respectively in Journal of Artificial Intelligence Research (JAIR 2022) [2] and the European Confernence in Information Retrieval (ECIR 2022) [1].…”

“…The second strategy, which we refer to as SAROS 𝑏 , is the same approach described in chapter 4, where model parameters are updated by minimizing a ranking loss over the same blocks of unclicked items followed by a clicked one using a gradient descent approach; with the difference that parameter updates are discarded for users who interact very little or a lot with the system. The results of this chapter were published at the Journal of Artificial Intelligence Research (JAIR) in 2021, [2] and the European Conference in Information Retrieval (ECIR) in 2022 [1].…”