Dynamic Time-aware Continual User Representation Learning

2025-04-23

AI Summary

Existing continual user representation learning methods neglect the impact of temporal progression on task evolution and item distribution shift during evaluation, leading to an overestimation of cross-task generalization capability. To address this, we propose the first time-aware continual learning evaluation paradigm, which more realistically simulates the dynamic evolution of user behavior and item distributions over time. We introduce DITTO—a novel framework integrating timestamp-driven task partitioning, Elastic Weight Consolidation (EWC), and dynamic prototype recalibration, augmented with distribution alignment to enable adaptive transfer of historical knowledge to the current drifting distribution. Evaluated under realistic temporal protocols, DITTO achieves a 12.7% improvement in Recall@10 over state-of-the-art methods, significantly mitigating catastrophic forgetting and performance degradation induced by cross-task distribution shifts.

Technology Category

• Machine Learning: Life-Long and Continual Learning
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Search and Optimization: Distributed Search

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Traditional user modeling (UM) approaches have primarily focused on designing models for a single specific task, but they face limitations in generalization and adaptability across various tasks. Recognizing these challenges, recent studies have shifted towards continual learning (CL)-based universal user representation learning aiming to develop a single model capable of handling multiple tasks. Despite advancements, existing methods are in fact evaluated under an unrealistic scenario that does not consider the passage of time as tasks progress, which overlooks newly emerged items that may change the item distribution of previous tasks. In this paper, we introduce a practical evaluation scenario on which CL-based universal user representation learning approaches should be evaluated, which takes into account the passage of time as tasks progress. Then, we propose a novel framework Dynamic Time-aware continual user representation learner, named DITTO, designed to alleviate catastrophic forgetting despite continuous shifts in item distribution, while also allowing the knowledge acquired from previous tasks to adapt to the current shifted item distribution. Through our extensive experiments, we demonstrate the superiority of DITTO over state-of-the-art methods under a practical evaluation scenario. Our source code is available at https://github.com/seungyoon-Choi/DITTO_official.

Problem

• Addressing limitations in generalization and adaptability across tasks in user modeling
• Evaluating continual learning methods under realistic time-aware scenarios with shifting item distributions
• Developing a framework to prevent catastrophic forgetting while adapting to distribution shifts

Innovation

• Dynamic Time-aware continual learning framework
• Alleviates catastrophic forgetting with shifting distributions
• Adapts previous knowledge to current item shifts