从轻量级卷积神经网络到脉冲神经网络：通过修剪脉冲压缩网络评估准确性与能量的权衡

Spiking Neural Networks (SNNs) are gaining attention as energy-efficient alternatives to Convolutional Neural Networks (CNNs), particularly in the context of edge intelligence. However, previous research has predominantly focused on large-scale models, leaving the design and evaluation of lightweight CNN-to-SNN pipelines underexplored. This study presents the first systematic benchmark of lightweight SNNs derived from compact CNN architectures, modeling activations with Leaky-Integrate-and-Fire (LIF) neurons and training using surrogate gradient descent in a unified framework. Spiking variants of ShuffleNet, SqueezeNet, MnasNet, and MixNet are constructed and evaluated on CIFAR-10, CIFAR-100, and TinyImageNet, measuring the trade-offs between accuracy and energy consumption.