Going Deeper With Directly-Trained Larger Spiking Neural Networks

Spiking neural networks (SNNs) are promising in a bio-plausible coding for spatio-temporal information and event-driven signal processing, which is very suited for energy-efficient implementation in neuromorphic hardware. However, the unique working mode of SNNs makes them more difficult to train than traditional networks. Currently, there are two main routes to explore the training of deep SNNs with high performance. The first is to convert a pre-trained ANN model to its SNN version, which usually requires a long coding window for convergence and cannot exploit the spatio-temporal features during training for solving temporal tasks. The other is to directly train SNNs in the spatio-temporal domain. But due to the binary spike activity of the firing function and the problem of gradient vanishing or explosion, current methods are restricted to shallow architectures and thereby difficult in harnessing large-scale datasets (e.g. ImageNet). To this end, we propose a threshold-dependent batch normalization (tdBN) method based on the emerging spatio-temporal backpropagation, termed "STBP-tdBN", enabling direct training of a very deep SNN and the efficient implementation of its inference on neuromorphic hardware. With the proposed method and elaborated shortcut connection, we significantly extend directly-trained SNNs from a shallow structure ( < 10 layer) to a very deep structure (50 layers). Furthermore, we theoretically analyze the effectiveness of our method based on "Block Dynamical Isometry" theory. Finally, we report superior accuracy results including 93.15 % on CIFAR-10, 67.8 % on DVS-CIFAR10, and 67.05% on ImageNet with very few timesteps. To our best knowledge, it's the first time to explore the directly-trained deep SNNs with high performance on ImageNet.

[1]  Matthew Cook,et al.  Fast-classifying, high-accuracy spiking deep networks through weight and threshold balancing , 2015, 2015 International Joint Conference on Neural Networks (IJCNN).

[2]  Chris Eliasmith,et al.  Spiking Deep Networks with LIF Neurons , 2015, ArXiv.

[3]  Steve B. Furber,et al.  Robustness of spiking Deep Belief Networks to noise and reduced bit precision of neuro-inspired hardware platforms , 2015, Front. Neurosci..

[4]  Kaushik Roy,et al.  Going Deeper in Spiking Neural Networks: VGG and Residual Architectures , 2018, Front. Neurosci..

[5]  Lei Deng,et al.  Spatio-Temporal Backpropagation for Training High-Performance Spiking Neural Networks , 2017, Front. Neurosci..

[6]  Hong Wang,et al.  Loihi: A Neuromorphic Manycore Processor with On-Chip Learning , 2018, IEEE Micro.

[7]  Andrew S. Cassidy,et al.  A million spiking-neuron integrated circuit with a scalable communication network and interface , 2014, Science.

[8]  Kaushik Roy,et al.  Enabling Spike-Based Backpropagation for Training Deep Neural Network Architectures , 2019, Frontiers in Neuroscience.

[9]  Kaiming He,et al.  Group Normalization , 2018, ECCV.

[10]  Garrick Orchard,et al.  SLAYER: Spike Layer Error Reassignment in Time , 2018, NeurIPS.

[11]  Ryad Benosman,et al.  HATS: Histograms of Averaged Time Surfaces for Robust Event-Based Object Classification , 2018, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.

[12]  Hong Yang,et al.  DART: Distribution Aware Retinal Transform for Event-Based Cameras , 2017, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[13]  Kaushik Roy,et al.  Enabling Deep Spiking Neural Networks with Hybrid Conversion and Spike Timing Dependent Backpropagation , 2020, ICLR.

[14]  Yuan Xie,et al.  Comparing SNNs and RNNs on Neuromorphic Vision Datasets: Similarities and Differences , 2020, Neural Networks.

[15]  Sergey Ioffe,et al.  Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift , 2015, ICML.

[16]  Tobi Delbrück,et al.  A Low Power, Fully Event-Based Gesture Recognition System , 2017, 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[17]  Wenrui Zhang,et al.  Hybrid Macro/Micro Level Backpropagation for Training Deep Spiking Neural Networks , 2018, NeurIPS.

[18]  Li Fei-Fei,et al.  ImageNet: A large-scale hierarchical image database , 2009, CVPR.

[19]  Jian Sun,et al.  Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[20]  Geoffrey E. Hinton,et al.  Layer Normalization , 2016, ArXiv.

[21]  Elisabetta Chicca,et al.  Efficient Processing of Spatio-Temporal Data Streams With Spiking Neural Networks , 2020, Frontiers in Neuroscience.

[22]  Luping Shi,et al.  CIFAR10-DVS: An Event-Stream Dataset for Object Classification , 2017, Front. Neurosci..

[23]  Kaushik Roy,et al.  RMP-SNN: Residual Membrane Potential Neuron for Enabling Deeper High-Accuracy and Low-Latency Spiking Neural Network , 2020, 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).

[24]  Gang Pan,et al.  Spiking Deep Residual Network , 2018, ArXiv.

[25]  Lei Deng,et al.  A Comprehensive and Modularized Statistical Framework for Gradient Norm Equality in Deep Neural Networks , 2020, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[26]  Tobi Delbrück,et al.  Training Deep Spiking Neural Networks Using Backpropagation , 2016, Front. Neurosci..

[27]  Kaushik Roy,et al.  Towards spike-based machine intelligence with neuromorphic computing , 2019, Nature.

[28]  Lei Deng,et al.  Direct Training for Spiking Neural Networks: Faster, Larger, Better , 2018, AAAI.