Is Neuromorphic MNIST Neuromorphic? Analyzing the Discriminative Power of Neuromorphic Datasets in the Time Domain

被引:19
作者
Iyer, Laxmi R. [1 ]
Chua, Yansong [1 ]
Li, Haizhou [1 ,2 ]
机构
[1] ASTAR, Inst Infocomms Res, Neuromorph Comp, Singapore, Singapore
[2] Huawei Technol Co Ltd, Shenzhen, Peoples R China
关键词
spiking neural network; spike timing dependent plasticity; N-MNIST dataset; neuromorphic benchmark; spike time coding; TEMPORAL STRUCTURE; NEURAL-NETWORK; SPIKE RATE; PRECISION; SYSTEM; BACKPROPAGATION; INFORMATION; CIRCUIT; NEURONS; DELAY;
D O I
10.3389/fnins.2021.608567
中图分类号
Q189 [神经科学];
学科分类号
071006 ;
摘要
A major characteristic of spiking neural networks (SNNs) over conventional artificial neural networks (ANNs) is their ability to spike, enabling them to use spike timing for coding and efficient computing. In this paper, we assess if neuromorphic datasets recorded from static images are able to evaluate the ability of SNNs to use spike timings in their calculations. We have analyzed N-MNIST, N-Caltech101 and DvsGesture along these lines, but focus our study on N-MNIST. First we evaluate if additional information is encoded in the time domain in a neuromorphic dataset. We show that an ANN trained with backpropagation on frame-based versions of N-MNIST and N-Caltech101 images achieve 99.23 and 78.01% accuracy. These are comparable to the state of the art-showing that an algorithm that purely works on spatial data can classify these datasets. Second we compare N-MNIST and DvsGesture on two STDP algorithms, RD-STDP, that can classify only spatial data, and STDP-tempotron that classifies spatiotemporal data. We demonstrate that RD-STDP performs very well on N-MNIST, while STDP-tempotron performs better on DvsGesture. Since DvsGesture has a temporal dimension, it requires STDP-tempotron, while N-MNIST can be adequately classified by an algorithm that works on spatial data alone. This shows that precise spike timings are not important in N-MNIST. N-MNIST does not, therefore, highlight the ability of SNNs to classify temporal data. The conclusions of this paper open the question-what dataset can evaluate SNN ability to classify temporal data?
引用
收藏
页数:21
相关论文
共 96 条
[1]   A Low Power, Fully Event-Based Gesture Recognition System [J].
Amir, Arnon ;
Taba, Brian ;
Berg, David ;
Melano, Timothy ;
McKinstry, Jeffrey ;
Di Nolfo, Carmelo ;
Nayak, Tapan ;
Andreopoulos, Alexander ;
Garreau, Guillaume ;
Mendoza, Marcela ;
Kusnitz, Jeff ;
Debole, Michael ;
Esser, Steve ;
Delbruck, Tobi ;
Flickner, Myron ;
Modha, Dharmendra .
30TH IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2017), 2017, :7388-7397
[2]  
[Anonymous], 2018, Advances in Neural Information Processing Systems
[3]   Neurogrid: A Mixed-Analog-Digital Multichip System for Large-Scale Neural Simulations [J].
Benjamin, Ben Varkey ;
Gao, Peiran ;
McQuinn, Emmett ;
Choudhary, Swadesh ;
Chandrasekaran, Anand R. ;
Bussat, Jean-Marie ;
Alvarez-Icaza, Rodrigo ;
Arthur, John V. ;
Merolla, Paul A. ;
Boahen, Kwabena .
PROCEEDINGS OF THE IEEE, 2014, 102 (05) :699-716
[4]   Graph-Based Object Classification for Neuromorphic Vision Sensing [J].
Bi, Yin ;
Chadha, Aaron ;
Abbas, Alhabib ;
Bourtsoulatze, Eirina ;
Andreopoulos, Yiannis .
2019 IEEE/CVF INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV 2019), 2019, :491-501
[5]  
Blake Randolph, 2005, Behav Cogn Neurosci Rev, V4, P21, DOI 10.1177/1534582305276839
[6]   A 240 x 180 130 dB 3 μs Latency Global Shutter Spatiotemporal Vision Sensor [J].
Brandli, Christian ;
Berner, Raphael ;
Yang, Minhao ;
Liu, Shih-Chii ;
Delbruck, Tobi .
IEEE JOURNAL OF SOLID-STATE CIRCUITS, 2014, 49 (10) :2333-2341
[7]   Philosophy of the Spike: Rate-Based vs. Spike-Based Theories of the Brain [J].
Brette, Romain .
FRONTIERS IN SYSTEMS NEUROSCIENCE, 2015, 9
[8]  
Butts DA, 2007, NATURE, V449, P92, DOI 10.1038/natureO6105
[9]   Asynchronous Convolutional Networks for Object Detection in Neuromorphic Cameras [J].
Cannici, Marco ;
Ciccone, Marco ;
Romanoni, Andrea ;
Matteucci, Matteo .
2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION WORKSHOPS (CVPRW 2019), 2019, :1656-1665
[10]   The Role of Conduction Delay in Creating Sensitivity to Interaural Time Differences [J].
Carr, Catherine ;
Ashida, Go ;
Wagner, Hermann ;
McColgan, Thomas ;
Kempter, Richard .
PHYSIOLOGY, PSYCHOACOUSTICS AND COGNITION IN NORMAL AND IMPAIRED HEARING, 2016, 894 :189-196