Rethinking residual connection in training large-scale spiking neural networks

Yudong Li, Yunlin Lei, Xu Yang

Abstract: Highlights•We analyze the merits and limitations of various residual connections and empirically demonstrate our analysis.•We abstract the concept of densely additive connection, extend to other topologies, and propose the DANet family.•We analyze the performance of the DANet family on various scales of datasets and demonstrate their advantages.