Henan Province Engineering Research Center of Spatial Information Processing, Kaifeng 475004, China.
College of Computer and Information Engineering, Henan University, Kaifeng 475004, China.
Comput Intell Neurosci. 2022 Oct 20;2022:1633946. doi: 10.1155/2022/1633946. eCollection 2022.
With the development of neuromorphic computing, more and more attention has been paid to a brain-inspired spiking neural network (SNN) because of its ultralow energy consumption and high-performance spatiotemporal information processing. Due to the discontinuity of the spiking neuronal activation function, it is still a difficult problem to train brain-inspired deep SNN directly, so SNN has not yet shown performance comparable to that of an artificial neural network. For this reason, the spike-based approximate backpropagation (SABP) algorithm and a general brain-inspired SNN framework are proposed in this paper. The combination of the two can be used for end-to-end direct training of brain-inspired deep SNN. Experiments show that compared with other spike-based methods of directly training SNN, the classification accuracy of this method is close to the best results on MNIST and CIFAR-10 datasets and achieves the best classification accuracy on sonar image target classification (SITC) of small sample datasets. Further analysis shows that compared with artificial neural networks, our brain-inspired SNN has great advantages in computational complexity and energy consumption in sonar target classification.
随着神经形态计算的发展，由于其超低的能耗和高性能的时空信息处理能力，人们越来越关注受大脑启发的尖峰神经网络 (SNN)。由于尖峰神经元激活函数的不连续性，直接训练受大脑启发的深度 SNN 仍然是一个难题，因此 SNN 尚未表现出与人工神经网络相当的性能。为此，本文提出了基于尖峰的近似反向传播 (SABP) 算法和通用的受大脑启发的 SNN 框架。这两者的结合可用于端到端直接训练受大脑启发的深度 SNN。实验表明，与其他直接训练 SNN 的基于尖峰的方法相比，该方法在 MNIST 和 CIFAR-10 数据集上的分类精度接近最佳结果，并在小样本数据集的声纳图像目标分类 (SITC) 上实现了最佳分类精度。进一步的分析表明，与人工神经网络相比，我们的受大脑启发的 SNN 在声纳目标分类的计算复杂度和能耗方面具有很大的优势。
