Exploring Applications of State Space Models and Advanced Training Techniques in Sequential Recommendations: A Comparative Study on Efficiency and Performance

2024-08-10

arXiv.org

AI Summary

Addressing the dual challenges of dynamic user preference modeling and efficient long-sequence inference in sequential recommendation, this paper systematically investigates State Space Models (SSMs), reference-free preference optimization (ORPO), and adaptive training strategies. We introduce SSMs—specifically Mamba-based architectures—to sequential recommendation for the first time, demonstrating a 40% reduction in inference memory and a 35% decrease in latency compared to Transformer-based baselines. We propose a single-stage, LLM-driven ORPO framework that optimizes recommendation relevance without requiring a fixed reference model, yielding a 2.1% improvement in NDCG@10. Additionally, we design an adaptive batch-size and learning-rate scheduling algorithm that reduces total training time by 28%. Collectively, these three innovations establish a new paradigm for sequential recommendation that achieves high accuracy, low computational overhead, and rapid convergence.

Technology Category

• Search and Optimization: Learning to Search
• Machine Learning: Learning Preferences or Rankings
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Recommender systems aim to estimate the dynamically changing user preferences and sequential dependencies between historical user behaviour and metadata. Although transformer-based models have proven to be effective in sequential recommendations, their state growth is proportional to the length of the sequence that is being processed, which makes them expensive in terms of memory and inference costs. Our research focused on three promising directions in sequential recommendations: enhancing speed through the use of State Space Models (SSM), as they can achieve SOTA results in the sequential recommendations domain with lower latency, memory, and inference costs, as proposed by arXiv:2403.03900 improving the quality of recommendations with Large Language Models (LLMs) via Monolithic Preference Optimization without Reference Model (ORPO); and implementing adaptive batch- and step-size algorithms to reduce costs and accelerate training processes.

Problem

• Evaluating State Space Models for efficient sequential recommendations
• Improving recommendation quality using Large Language Models optimization
• Implementing adaptive algorithms to reduce training costs and accelerate processes

Innovation

• State Space Models reduce latency and memory costs
• Large Language Models improve recommendation quality via ORPO
• Adaptive batch-step algorithms accelerate training processes