DS-AL: A Dual-Stream Analytic Learning for Exemplar-Free Class-Incremental Learning
About
Class-incremental learning (CIL) under an exemplar-free constraint has presented a significant challenge. Existing methods adhering to this constraint are prone to catastrophic forgetting, far more so than replay-based techniques that retain access to past samples. In this paper, to solve the exemplar-free CIL problem, we propose a Dual-Stream Analytic Learning (DS-AL) approach. The DS-AL contains a main stream offering an analytical (i.e., closed-form) linear solution, and a compensation stream improving the inherent under-fitting limitation due to adopting linear mapping. The main stream redefines the CIL problem into a Concatenated Recursive Least Squares (C-RLS) task, allowing an equivalence between the CIL and its joint-learning counterpart. The compensation stream is governed by a Dual-Activation Compensation (DAC) module. This module re-activates the embedding with a different activation function from the main stream one, and seeks fitting compensation by projecting the embedding to the null space of the main stream's linear mapping. Empirical results demonstrate that the DS-AL, despite being an exemplar-free technique, delivers performance comparable with or better than that of replay-based methods across various datasets, including CIFAR-100, ImageNet-100 and ImageNet-Full. Additionally, the C-RLS' equivalent property allows the DS-AL to execute CIL in a phase-invariant manner. This is evidenced by a never-before-seen 500-phase CIL ImageNet task, which performs on a level identical to a 5-phase one. Our codes are available at https://github.com/ZHUANGHP/Analytic-continual-learning.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Class-incremental learning | ImageNet-100 B=50, C=10 1.0 | Avg Incremental Acc75.2 | 42 | |
| Incremental Learning | CIFAR-100 (test) | Accuracy (S9)61.4 | 26 | |
| Class-incremental learning | CIFAR-100 B50Inc10 | Accuracy (t=5)0.5843 | 24 | |
| Exemplar-Free Class-Incremental Learning | TinyImageNet T=10 | Alast33.6 | 16 | |
| Incremental Learning | ImageNet subset (test) | Last Acc68 | 14 | |
| Incremental Learning | ImageNet-100 B50Inc10 (T=5) (test) | Average Accuracy75.2 | 12 | |
| Incremental Learning | CIFAR-100 B50Inc10 (T=5) | Accuracy66.54 | 12 | |
| Incremental Learning | CIFAR-100 B50Inc5 (T=10) | Avg Accuracy66.46 | 11 | |
| Incremental Learning | CIFAR-100 B40Inc3 (T=20) | Avg Accuracy42.05 | 11 | |
| Exemplar-Free Class-Incremental Learning | Imagenet Subset T=20 (test) | Alast36.7 | 10 |