ann

Hybrid ANN-SNN Pipeline Achieves 99.09% Accuracy on ImageNet via Local Plasticity

A hybrid artificial-spiking neural network pipeline, coupling a pretrained EfficientNet encoder with a CoLaNET spiking classifier, achieves 99.09% accuracy on a 64-class ImageNet benchmark. The approach uses rate-coding to convert activations to spike trains and trains the SNN classifier with local, biologically inspired learning rules, bypassing end-to-end gradient propagation.