Towards Better Plasticity-Stability Trade-off in Incremental Learning: A Simple Linear Connector
This addresses a key problem in incremental learning for applications where storing past data is not feasible, offering a novel approach to improve trade-offs, though it appears incremental in nature.
The paper tackles the plasticity-stability dilemma in incremental learning by proposing a method that uses mode connectivity in loss landscapes to balance learning new tasks with retaining old knowledge without storing previous samples, achieving improvements such as 6.02% higher accuracy on 10-split-CIFAR-100 and 6.33% on TinyImageNet.
Plasticity-stability dilemma is a main problem for incremental learning, where plasticity is referring to the ability to learn new knowledge, and stability retains the knowledge of previous tasks. Many methods tackle this problem by storing previous samples, while in some applications, training data from previous tasks cannot be legally stored. In this work, we propose to employ mode connectivity in loss landscapes to achieve better plasticity-stability trade-off without any previous samples. We give an analysis of why and how to connect two independently optimized optima of networks, null-space projection for previous tasks and simple SGD for the current task, can attain a meaningful balance between preserving already learned knowledge and granting sufficient flexibility for learning a new task. This analysis of mode connectivity also provides us a new perspective and technology to control the trade-off between plasticity and stability. We evaluate the proposed method on several benchmark datasets. The results indicate our simple method can achieve notable improvement, and perform well on both the past and current tasks. On 10-split-CIFAR-100 task, our method achieves 79.79% accuracy, which is 6.02% higher. Our method also achieves 6.33% higher accuracy on TinyImageNet. Code is available at https://github.com/lingl1024/Connector.