Ferret: An Efficient Online Continual Learning Framework under Varying Memory Constraints

2025-03-15

AI Summary

To address the challenge of balancing real-time responsiveness and accuracy in online continual learning (OCL) under high-frequency data streams and dynamic memory constraints, this paper proposes a fine-grained pipelined parallel framework coupled with an iterative gradient compensation algorithm. The framework enables memory-adaptive model partitioning and dynamic pipeline scheduling, while the algorithm mitigates gradient staleness. Together, they jointly overcome the dual bottlenecks of resource volatility and stale gradients. Extensive evaluation across 20 benchmark datasets and five state-of-the-art OCL algorithms demonstrates that our method reduces peak memory consumption by up to 3.7× while preserving comparable online accuracy. Moreover, it consistently achieves superior performance across diverse memory budgets, significantly enhancing learning efficiency and robustness in resource-constrained environments.

Technology Category

• Machine Learning: Life-Long and Continual Learning
• Search and Optimization: Learning to Search
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

In the realm of high-frequency data streams, achieving real-time learning within varying memory constraints is paramount. This paper presents Ferret, a comprehensive framework designed to enhance online accuracy of Online Continual Learning (OCL) algorithms while dynamically adapting to varying memory budgets. Ferret employs a fine-grained pipeline parallelism strategy combined with an iterative gradient compensation algorithm, ensuring seamless handling of high-frequency data with minimal latency, and effectively counteracting the challenge of stale gradients in parallel training. To adapt to varying memory budgets, its automated model partitioning and pipeline planning optimizes performance regardless of memory limitations. Extensive experiments across 20 benchmarks and 5 integrated OCL algorithms show Ferret's remarkable efficiency, achieving up to 3.7$ imes$ lower memory overhead to reach the same online accuracy compared to competing methods. Furthermore, Ferret consistently outperforms these methods across diverse memory budgets, underscoring its superior adaptability. These findings position Ferret as a premier solution for efficient and adaptive OCL framework in real-time environments.

Problem

• Enhance online accuracy in high-frequency data streams
• Adapt to varying memory constraints dynamically
• Minimize latency and counteract stale gradients

Innovation

• Fine-grained pipeline parallelism strategy
• Iterative gradient compensation algorithm
• Automated model partitioning and pipeline planning