Active Transfer Learning Strategy for Spiking Neural Networks

In recent years, Spiking Neural Networks (SNNs) have attracted considerable attention due to their low power consumption and high robustness. However, SNN models are inherently difficult to train and require a large amount of labeled data to achieve satisfactory performance. To reduce the reliance on large labeled datasets, active learning and transfer learning methods have been introduced in the field of Artificial Neural Networks (ANNs). However, their application to SNNs remains in the early stages of exploration. In this paper, we propose ATSNN, an active transfer learning method for SNNs, which effectively addresses both domain adaptation and the scarcity of labeled samples during model training. First, we design a joint domain adaptation loss function to enhance transfer learning within the SNN architecture. Second, an active learning strategy is employed to provide higher-quality labeled data for the target domain, further improving transfer learning performance. Extensive experiments on multiple datasets demonstrate that ATSNN improves SNN performance in the target domain while requiring fewer labeled samples.