Conversion of Continuous-Valued Deep Networks to Efficient Event-Driven Networks for Image Classification.

Bodo Rueckauer,Shih-Chii Liu,Yuhuang Hu,Iulia-Alexandra Lungu,Michael Pfeiffer

doi:10.3389/fnins.2017.00682

Abstract

Spiking neural networks (SNNs) can potentially offer an efficient way of doing inference because the neurons in the networks are sparsely activated and computations are event-driven. Previous work showed that simple continuous-valued deep Convolutional Neural Networks (CNNs) can be converted into accurate spiking equivalents. These networks did not include certain common operations such as max-pooling, softmax, batch-normalization and Inception-modules. This paper presents spiking equivalents of these operations therefore allowing conversion of nearly arbitrary CNN architectures. We show conversion of popular CNN architectures, including VGG-16 and Inception-v3, into SNNs that produce the best results reported to date on MNIST, CIFAR-10 and the challenging ImageNet dataset. SNNs can trade off classification error rate against the number of available operations whereas deep continuous-valued neural networks require a fixed number of operations to achieve their classification error rate. From the examples of LeNet for MNIST and BinaryNet for CIFAR-10, we show that with an increase in error rate of a few percentage points, the SNNs can achieve more than 2x reductions in operations compared to the original CNNs. This highlights the potential of SNNs in particular when deployed on power-efficient neuromorphic spiking neuron chips, for use in embedded applications.

Highlights

Deep Artificial Neural Network (ANN) architectures such as GoogLeNet (Szegedy et al, 2015) and VGG-16 (Simonyan and Zisserman, 2014) have successfully pushed the state-of-the-art classification error rates to new levels on challenging computer vision benchmarks like ImageNet (Russakovsky et al, 2015)
There are two ways of improving the classification error rate of an Spiking Neural Networks (SNNs) obtained via conversion: (1) training a better ANN before conversion, and (2) improving the conversion by eliminating approximation errors of the SNN
By implementing SNN-compatible versions of common ANN Convolutional Neural Networks (CNNs) features such as max pooling, softmax, batch normalization, biases and Inception modules, we allow a larger class of CNNs including VGG-16 and GoogLeNet Inception-V3 to be converted into SNNs

Summary

Introduction

Deep Artificial Neural Network (ANN) architectures such as GoogLeNet (Szegedy et al, 2015) and VGG-16 (Simonyan and Zisserman, 2014) have successfully pushed the state-of-the-art classification error rates to new levels on challenging computer vision benchmarks like ImageNet (Russakovsky et al, 2015). Inference in such very large networks, i.e., classification of an ImageNet frame, requires substantial computational and energy costs, limiting their use in mobile and embedded applications. SNNs could play an important role in supporting, or in some cases replacing deep ANNs in tasks where fast and efficient classification in real-time is crucial, such as detection of objects in larger and moving scenes, tracking tasks, or activity recognition (Hu et al, 2016)

Methods

Results

Conclusion