Recommender systems: when memory matters

TL;DR

This paper investigates the impact of long memory in user interactions on the learnability of sequential recommender systems, proposing an online user-specific update algorithm and demonstrating performance gains through empirical evaluation.

Contribution

It introduces a novel online algorithm that updates model parameters user-by-user considering long memory effects, improving large-scale recommender system performance.

Findings

Filtering users by long memory improves MAP and NDCG scores.

Long memory consideration enhances learning efficiency in large-scale systems.

Empirical results validate the effectiveness of the proposed approach.

Abstract

In this paper, we study the effect of long memory in the learnability of a sequential recommender system including users' implicit feedback. We propose an online algorithm, where model parameters are updated user per user over blocks of items constituted by a sequence of unclicked items followed by a clicked one. We illustrate through thorough empirical evaluations that filtering users with respect to the degree of long memory contained in their interactions with the system allows to substantially gain in performance with respect to MAP and NDCG, especially in the context of training large-scale Recommender Systems.