SNV applies Shapley values to rank and freeze neurons in neural nets for buffer-free continual learning, reporting accuracy gains of +2.88% (class-incremental) and +6.46% (task-incremental) over baselines on ImageNet-1k.
AISTATS , year =
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Shapley Neuron Values for Continual Learning: Which Neurons Matter Most?
SNV applies Shapley values to rank and freeze neurons in neural nets for buffer-free continual learning, reporting accuracy gains of +2.88% (class-incremental) and +6.46% (task-incremental) over baselines on ImageNet-1k.