What problem does this paper attempt to address?

The problem that this paper attempts to solve is the technical challenges encountered in the training process of existing Spiking Neural Networks (SNNs), especially how to avoid the local minimum problem during the training process. Due to their energy - efficiency advantages, SNNs are regarded as a potential solution to the high - energy - consumption problem of current Artificial Neural Networks (ANNs). However, effective SNNs training methods still face great challenges. In particular, although surrogate - gradient - based methods provide a feasible solution, these methods are prone to make SNNs fall into local minima, which limits their performance and application scope. To solve these problems, Chen's research team introduced a new training method - Chaotic Spiking Backpropagation (CSBP). Inspired by the chaotic dynamics observed in the brain's learning process, this method simulates brain - like chaotic dynamics by introducing specific terms in the loss function, and utilizes its ergodicity and pseudo - random properties to enhance the effectiveness and robustness of SNNs learning. The CSBP method not only improves the training effect of SNNs, but also provides a new perspective for understanding the learning mechanism of the human brain. Specifically, CSBP has the following key features: 1. **Intrinsically integrates chaotic dynamics inspired by the animal brain**: This enables CSBP to simulate dynamic behaviors closer to real biological nervous systems. 2. **Theoretically guarantees global search ability and convergence**: It helps to avoid the local minimum problem common in traditional methods. 3. **Surpasses the existing state - of - the - art methods**: It performs well on a variety of datasets, including neuromorphic datasets (such as DVS - CIFAR10 and DVS - Gesture) and large - scale static datasets (such as ImageNet and CIFAR100). 4. **Functions as an additional module or plug - in**: It can enhance the performance of any existing method by simply introducing a term without changing the parameters of the original method. 5. **Is robust to initial conditions and hyper - parameters**: It effectively alleviates the local minimum trap problem common in existing methods. In summary, the main objective of this paper is to propose a new SNNs training method, improve the learning efficiency and robustness of SNNs by introducing chaotic dynamics, and at the same time provide new theoretical support for understanding the brain learning mechanism.

A chaotic spiking backpropagation approach to brain learning

Brain-inspired Chaotic Spiking Backpropagation

Backpropagation with Sparsity Regularization for Spiking Neural Network Learning.

Toward Efficient Processing and Learning with Spikes: New Approaches for Multispike Learning

Biologically Inspired Structure Learning with Reverse Knowledge Distillation for Spiking Neural Networks

A Brain-Inspired Spiking Neural Network Model with Temporal Encoding and Learning

BioLeaF: A Bio-plausible Learning Framework for Training of Spiking Neural Networks

Training Spiking Neural Networks Using Lessons From Deep Learning

A solution to the learning dilemma for recurrent networks of spiking neurons

BrainScale: Enabling Scalable Online Learning in Spiking Neural Networks

Advanced Computing and Related Applications Leveraging Brain-inspired Spiking Neural Networks

Toward spike-based stochastic neural computing

Temporal Surrogate Back-propagation for Spiking Neural Networks

Spike-driven Multi-Scale Learning with Hybrid Mechanisms of Spiking Dendrites.

A temporally and spatially local spike-based backpropagation algorithm to enable training in hardware

Effect Exploration of the Spiking Neuron Property and Heterogeneity on Brain-Inspired Context-Dependent Learning

Decoding Spiking Mechanism with Dynamic Learning on Neuron Population

Comprehensive Online Training and Deployment for Spiking Neural Networks

MAP-SNN: Mapping spike activities with multiplicity, adaptability, and plasticity into bio-plausible spiking neural networks

Training Spiking Neural Networks with Event-driven Backpropagation

Research Advances and New Paradigms for Biology-inspired Spiking Neural Networks