A Faster Path to Continual Learning

2026-04-13

AI Summary

This work addresses the high computational cost and low training efficiency of optimization methods in continual learning by proposing the C-Flat Turbo optimizer. It reveals, for the first time, the directional invariance inherent in first-order gradient flatness and leverages inter-task gradient stability to design an adaptive, linearly scheduled triggering mechanism that effectively reduces redundant gradient computations. Without compromising model performance, C-Flat Turbo achieves 1.0–1.25× training speedup across multiple continual learning benchmarks while maintaining comparable or superior accuracy, thereby significantly enhancing training efficiency.

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

Continual Learning (CL) aims to train neural networks on a dynamic stream of tasks without forgetting previously learned knowledge. Among optimization-based approaches, C-Flat has emerged as a promising solution due to its plug-and-play nature and its ability to encourage uniformly low-loss regions for both new and old tasks. However, C-Flat requires three additional gradient computations per iteration, imposing substantial overhead on the optimization process. In this work, we propose C-Flat Turbo, a faster yet stronger optimizer that significantly reduces the training cost. We show that the gradients associated with first-order flatness contain direction-invariant components relative to the proxy-model gradients, enabling us to skip redundant gradient computations in the perturbed ascent steps. Moreover, we observe that these flatness-promoting gradients progressively stabilize across tasks, which motivates a linear scheduling strategy with an adaptive trigger to allocate larger turbo steps for later tasks. Experiments show that C-Flat Turbo is 1.0$\times$ to 1.25$\times$ faster than C-Flat across a wide range of CL methods, while achieving comparable or even improved accuracy.

Problem

• Continual Learning
• C-Flat
• gradient computation
• training overhead
• optimization efficiency

Innovation

• Continual Learning
• C-Flat Turbo
• gradient efficiency
• flatness optimization
• adaptive scheduling