Time-aware top-N recommender systems based on reinforcement learning

• Grant number: 21/14591-7
• Support Opportunities: Scholarships in Brazil - Doctorate
• Start date: July 01, 2022
• End date: October 25, 2026
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: Tiago Agostinho de Almeida
• Grantee: Pedro Reis Pires
• Host Institution: Centro de Ciências em Gestão e Tecnologia (CCGT). Universidade Federal de São Carlos (UFSCAR). Campus de Sorocaba. Sorocaba , SP, Brazil
• Associated scholarship(s): 24/15919-4 - Time-aware exploration and state representation for incremental recommender systems, BE.EP.DR

Abstract

With the constant popularization of technology, recommendation systems have become increasingly important within digital media. Its main purpose is to recommend a subset of items relevant to a specific user, helping them to discover new interests. Since the beginning of the field, a static and non-incremental approach has been common, in which algorithms are trained with a fixed database, captured in the past. However, the practical recommendation scenario operates sequentially: the system generates recommendations to the user, who immediately provides feedback. Recent studies in the area are developing models that take advantage of this continuous feature of the problem. Based on recent reinforcement learning strategies, these methods learn in an incremental manner, thus generating a recommending agent that automatically adapts to the users' interests over time. As this area of research is quite recent, it is still maturing and has many open challenges. For example, models capable of generating top-N recommendations, i.e., a list of ordered items, are practically scarce. Additionally, although reinforcement learning has, by definition, sequential temporal knowledge of interactions, little is known about the explicit use of time during training, as is done in time-aware recommenders. This research project seeks to study these open questions and propose new recommendation models based on reinforcement learning for top-N recommendation tasks. The main strategy will be the balance between exploration and diversity of the list of recommended items, in addition to the consumption of temporal attributes to increase the agents' knowledge. (AU)

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

PIRES, PEDRO R.; ALMEIDN, TIAGO A.; XAVIER-JUNIOR, JC; RIOS, RA. Interact2Vec: Neural Item and User Embedding for Collaborative Filtering. INTELLIGENT SYSTEMS, PT II, v. 13654, p. 16-pg., 2022-01-01. (21/14591-7)

PIRES, PEDRO R.; ALMEIDA, TIAGO A.. Interact2Vec-An efficient neural network-based model for simultaneously learning users and items embeddings in recommender systems. APPLIED SOFT COMPUTING, v. 181, p. 17-pg., 2025-09-01. (21/14591-7)

PIRES, PEDRO R.; SEREICIKAS, RAFAEL T.; AZEVEDO, GREGORIO F.; ALMEIDA, TIAGO A.. Collaborative filtering through weighted similarities of user and item embeddings. 40TH ANNUAL ACM SYMPOSIUM ON APPLIED COMPUTING, v. N/A, p. 8-pg., 2025-01-01. (21/14591-7, 23/00158-5)