This paper introduces a novel approach to sequential recommendation systems, a critical component of online platforms and services. The authors propose the application of the Fisher-Merging method to Sequential Recommendation for the first time, aiming to address the challenges of data sparsity due to limited user-item interactions. The method merges the parameters of multiple models, ensuring robust fine-tuning and improved overall performance. The paper demonstrates the effectiveness of this approach through extensive experiments, suggesting its potential to advance the state-of-the-art in sequential learning and recommendation systems.
Publication date: July 5, 2023
Project Page: N/A
Paper: https://arxiv.org/pdf/2307.05476.pdf
