Shapley Neuron Values for Continual Learning: Which Neurons Matter Most?
For researchers in continual learning, SNV provides a principled, buffer-free approach that outperforms existing methods on ImageNet-1k.
The paper tackles catastrophic forgetting in continual learning by introducing Shapley Neuron Valuation (SNV), a game-theoretic method to quantify neuron importance and selectively freeze important neurons. On ImageNet-1k, SNV improves accuracy by +2.88% in class incremental learning and +6.46% in task incremental learning over the second-best baseline.
Continual learning enables neural networks to learn tasks sequentially without forgetting previously acquired knowledge. However, neural networks suffer from catastrophic forgetting, where learning new tasks degrades performance on earlier ones. We address this problem with Shapley Neuron Valuation (SNV), a principled framework that quantifies Neuron importance in continual learning, grounded in cooperative game theory. SNV selectively freezes important Neurons while keeping others plastic, enabling buffer-free continual learning without expanding architecture. Experiments on ImageNet-1k show that SNV consistently outperforms existing buffer-free methods. In particular, SNV improves accuracy by +2.88% in the class incremental learning and +6.46% in the task incremental learning scenarios compared to the second baseline.