Hybrid ANN-SNN Pipeline with Local Plasticity
Researchers have developed a novel hybrid pipeline that combines Artificial Neural Networks (ANNs) with Spiking Neural Networks (SNNs) to achieve high performance on image classification tasks. The system utilizes a pretrained EfficientNet encoder to generate spike trains, which are then fed into a CoLaNET spiking classifier trained with local, biologically inspired learning rules. This method bypasses the need for end-to-end gradient propagation and has demonstrated 99.09% accuracy on a 64-class ImageNet benchmark, matching the performance of traditional deep networks. AI
IMPACT This research offers a more biologically plausible and efficient framework for adapting powerful pretrained models to downstream tasks, potentially influencing future SNN development.