Abstract
Event-driven mode of computation provides SNNs with potential to bridge the gap between excellent performance and computational load of deep neural networks. However, SNNs are difficult to train because of the discontinuity of spike signals. This paper proposes an efficient framework for CNN-to-SNN conversion, which converts pre-trained convolution neural networks (CNNs) into corresponding spiking equivalents. Different from previous work, this paper focuses on the conversion of deep CNN architectures, such as Inception and ResNet. As networks in conversion are rate-encoding, a novel weight normalization method is employed to approximate the spiking rates of SNNs to the activations of CNNs. And, inspired from homeostatic plasticity in neural system, a compensation approach is introduced to reduce the deterioration of spiking rates at deep layers and accelerate the inference of SNNs. Experimental results on CIFAR dataset show that the SNNs built by the conversion framework achieve better performance than those trained with spike-based algorithms. In particular, the accuracy gap between converted SNNs and original CNNs is further reduced, which is helpful for large-scale employment of spiking networks.
This is a preview of subscription content, log in via an institution.
Buying options
Tax calculation will be finalised at checkout
Purchases are for personal use only
Learn about institutional subscriptionsReferences
Neil, D., Liu, S.C.: Minitaur, an event-driven FPGA-based spiking network accelerator. IEEE Trans. Very Large Scale Integr. (VLSI) Syst. 22(12), 2621–2628 (2014)
Liu, S.C., Delbruck, T., Indiveri, G., Douglas, R., Whatley, A.: Event-Based Neuromorphic Systems. Wiley, Hoboken (2015)
Merolla, P.A., et al.: A million spiking-neuron integrated circuit with a scalable communication network and interface. Science 345(6197), 668–673 (2014)
Furber, S.B., Galluppi, F., Temple, S., Plana, L.A.: The SpiNNaker project. Proc. IEEE 102(5), 652–665 (2014)
Diehl, P.U., Cook, M.: Unsupervised learning of digit recognition using spike-timing-dependent plasticity. Front. Comput. Neurosci. 9, 99 (2015)
Pérez-Carrasco, J.A., et al.: Mapping from frame-driven to frame-free event-driven vision systems by low-rate rate coding and coincidence processing–application to feedforward ConvNets. IEEE Trans. Pattern Anal. Mach. Intell. 35(11), 2706–2719 (2013)
Cao, Y., Chen, Y., Khosla, D.: Spiking deep convolutional neural networks for energy-efficient object recognition. Int. J. Comput. Vis. 113(1), 54–66 (2015)
Diehl, P.U., Neil, D., Binas, J., Cook, M., Liu, S.C., Pfeiffer, M.: Fast-classifying, high-accuracy spiking deep networks through weight and threshold balancing. In: 2015 International Joint Conference on Neural Networks (IJCNN), pp. 1–8. IEEE (2015)
Rueckauer, B., Lungu, I.A., Hu, Y., Pfeiffer, M., Liu, S.C.: Conversion of continuous-valued deep networks to efficient event-driven networks for image classification. Front. Neurosci. 11, 682 (2017)
Hu, Y., Tang, H., Wang, Y., Pan, G.: Spiking deep residual network. arXiv preprint arXiv:1805.01352 (2018)
Szegedy, C., et al.: Going deeper with convolutions. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1–9 (2015)
Szegedy, C., Vanhoucke, V., Ioffe, S., Shlens, J., Wojna, Z.: Rethinking the inception architecture for computer vision. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2818–2826 (2016)
Szegedy, C., Ioffe, S., Vanhoucke, V., Alemi, A.A.: Inception-v4, Inception-ResNet and the impact of residual connections on learning. In: AAAI, vol. 4, p. 12 (2017)
He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 770–778 (2016)
Fernandes, D., Carvalho, A.L.: Mechanisms of homeostatic plasticity in the excitatory synapse. J. Neurochem. 139(6), 973–996 (2016)
Krizhevsky, A., Hinton, G.: Learning multiple layers of features from tiny images. Technical report, Citeseer (2009)
Salimans, T., Kingma, D.P.: Weight normalization: a simple reparameterization to accelerate training of deep neural networks. In: Advances in Neural Information Processing Systems, pp. 901–909 (2016)
Simonyan, K., Zisserman, A.: Very deep convolutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556 (2014)
Turrigiano, G.G., Nelson, S.B.: Homeostatic plasticity in the developing nervous system. Nat. Rev. Neurosci. 5(2), 97 (2004)
Miner, D., Triesch, J.: Plasticity-driven self-organization under topological constraints accounts for non-random features of cortical synaptic wiring. PLoS Comput. Biol. 12(2), e1004759 (2016)
Deng, J., Dong, W., Socher, R., Li, L.J., Li, K., Fei-Fei, L.: ImageNet: a large-scale hierarchical image database. In: 2009 IEEE Conference on Computer Vision and Pattern Recognition, pp. 248–255. IEEE (2009)
Esser, S., et al.: Convolutional networks for fast, energy-efficient neuromorphic computing. Preprint on ArXiv http://arxiv.org/abs/1603.08270 (2016). Accessed 27 2016
Acknowledgments
This study was partly supported by the National Natural Science Foundation of China (No. 41571402), the Science Fund for Creative Research Groups of the National Natural Science Foundation of China (No. 61221003), and National Key R&D Program of China (No. 2018YF-B0505000).
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2019 Springer Nature Switzerland AG
About this paper
Cite this paper
Xing, F., Yuan, Y., Huo, H., Fang, T. (2019). Homeostasis-Based CNN-to-SNN Conversion of Inception and Residual Architectures. In: Gedeon, T., Wong, K., Lee, M. (eds) Neural Information Processing. ICONIP 2019. Lecture Notes in Computer Science(), vol 11955. Springer, Cham. https://doi.org/10.1007/978-3-030-36718-3_15
Download citation
DOI: https://doi.org/10.1007/978-3-030-36718-3_15
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-36717-6
Online ISBN: 978-3-030-36718-3
eBook Packages: Computer ScienceComputer Science (R0)