A Deep Unsupervised Feature Learning Spiking Neural Network with\n Binarized Classification Layers for EMNIST Classification using SpykeFlow

2020/02/26 by Ruthvik Vaila, Vaila, Ruthvik, John Chiasson +3
Computer Science · Engineering · Neuroscience · #Advanced Memory and Neural Computing #FOS: Biological sciences #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #Machine Learning (cs.LG) #Neural Networks and Reservoir Computing #Neural and Evolutionary Computing (cs.NE) #Neural dynamics and brain function #Neurons and Cognition (q-bio.NC) #electronic engineering #information engineering

paper · pdf · doi:10.48550/arxiv.2002.11843

openalex publication_date 2020/02/26 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

End user AI is trained on large server farms with data collected from the\nusers. With ever increasing demand for IOT devices, there is a need for deep\nlearning approaches that can be implemented (at the edge) in an energy\nefficient manner. In this work we approach this using spiking neural networks.\nThe unsupervised learning technique of spike timing dependent plasticity (STDP)\nusing binary activations are used to extract features from spiking input data.\nGradient descent (backpropagation) is used only on the output layer to perform\nthe training for classification. The accuracies obtained for the balanced\nEMNIST data set compare favorably with other approaches. The effect of\nstochastic gradient descent (SGD) approximations on learning capabilities of\nour network are also explored.\n

Citations

• Matplotlib: A 2D Graphics Environment
• Gradient-based learning applied to document recognition
• Optimization Methods for Large-Scale Machine Learning
• Theory and Tools for the Conversion of Analog to Spiking Convolutional Neural Networks
• EMNIST: an extension of MNIST to handwritten letters
• Hybrid Macro/Micro Level Backpropagation for Training Deep Spiking Neural Networks
• PyTorch: An Imperative Style, High-Performance Deep Learning Library

Related

• Deep Spiking Neural Network with Spike Count based Learning Rule
• Deep Convolutional Spiking Neural Networks for Image Classification
• Exploring Spiking Neural Networks for Binary Classification in Multivariate Time Series at the Edge
• Efficient Memristive Spiking Neural Networks Architecture with Supervised In-Situ STDP Method
• Enabling Deep Spiking Neural Networks with Hybrid Conversion and Spike Timing Dependent Backpropagation