SAFA-SNN combines sparsity-aware spike dynamics and orthogonal subspace projection in spiking networks to achieve on-device few-shot class-incremental learning with lower energy use and reduced forgetting than prior baselines.
At the non-differentiable pointu= 0, the true gradient does not exist
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
SAFA-SNN: Sparsity-Aware On-Device Few-Shot Class-Incremental Learning with Fast-Adaptive Structure of Spiking Neural Network
SAFA-SNN combines sparsity-aware spike dynamics and orthogonal subspace projection in spiking networks to achieve on-device few-shot class-incremental learning with lower energy use and reduced forgetting than prior baselines.