A Highly Effective and Robust Membrane Potential-Driven Supervised Learning Method for Spiking Neurons

Abstract

Spiking neurons are becoming increasingly popular owing to their biological plausibility and promising computational properties. Unlike traditional rate-based neural models, spiking neurons encode information in the temporal patterns of the transmitted spike trains, which makes them more suitable for processing spatiotemporal information. One of the fundamental computations of spiking neurons is to transform streams of input spike trains into precisely timed firing activity. However, the existing learning methods, used to realize such computation, often result in relatively low accuracy performance and poor robustness to noise. In order to address these limitations, we propose a novel highly effective and robust membrane potential-driven supervised learning (MemPo-Learn) method, which enables the trained neurons to generate desired spike trains with higher precision, higher efficiency, and better noise robustness than the current state-of-the-art spiking neuron learning methods. While the traditional spike-driven learning methods use an error function based on the difference between the actual and desired output spike trains, the proposed MemPo-Learn method employs an error function based on the difference between the output neuron membrane potential and its firing threshold. The efficiency of the proposed learning method is further improved through the introduction of an adaptive strategy, called skip scan training strategy, that selectively identifies the time steps when to apply weight adjustment. The proposed strategy enables the MemPo-Learn method to effectively and efficiently learn the desired output spike train even when much smaller time steps are used. In addition, the learning rule of MemPo-Learn is improved further to help mitigate the impact of the input noise on the timing accuracy and reliability of the neuron firing dynamics. The proposed learning method is thoroughly evaluated on synthetic data and is further demonstrated on real-world classification tasks. Experimental results show that the proposed method can achieve high learning accuracy with a significant improvement in learning time and better robustness to different types of noise.

Spiking Neural Network Based on Cusp Catastrophe Theory

“…As mentioned in Section 1, the key aspect of most systems is power consumption and strategies for minimizing it. At the same time, the implementation of spiking networks is an area of interest of many IT corporations, as well as a subject of numerous scientific research dedicated to the problem of training [9,20] or hardware implementation [6,19]. Models of neurons shown in Figure 2 were developed for hardware implementation using modern nanometer CMOS technologies.…”

SpiFoG: an efficient supervised learning algorithm for the network of spiking neurons

“…J. Wang et al 32 proposed OSNN algorithm having adaptive structure based supervised learning ability. There exist membrane potential driven learning algorithms proposed for SNN such as MPD-AL 33 , EMPD 34 , and MemPo-Learn 35 . In the development of MPD-AL 33 algorithm, two approaches were used namely firing less spikes than desired and firing more spikes than desired.…”

“…In 34 , two process namely at desired output times and at undesired output times were used. The error function proposed in 35 is formed by taking difference between membrane potential of output neuron and the firing threshold of its own. The ability of evolutionary strategies (ES) to work directly with real numbers without complex encoding technique drives researchers towards the investigation of ES to train SNN.…”

An Efficient and Perceptually Motivated Auditory Neural Encoding and Decoding Algorithm for Spiking Neural Networks

“…On the other hand, a dataset has to be complex enough such that it simulates a real-world problem. There are some datasets that support the learning of temporal patterns (Zhang et al, 2017(Zhang et al, , 2018(Zhang et al, , 2019Wu et al, 2018a), whereby each pattern contains only a single label, such as a sound event or an isolated word. Such datasets are much simpler than those in deep learning studies (Graves et al, 2006), whereby a temporal pattern involves a sequence of labels, such as continuous speech.…”