Advances in Design and Application of Spiking Neural Networks

Abstract

This paper presents new findings in the design and application of biologically plausible neural networks based on spiking neuron models, which represent a more plausible model of real biological neurons where time is considered as an important feature for information encoding and processing in the brain. The design approach consists of an evolutionary strategy based supervised training algorithm, newly developed by the authors, and the use of different biologically plausible neuronal models. A dynamic synapse (DS) based neuron model, a biologically more detailed model, and the spike response model (SRM) are investigated in order to demonstrate the efficacy of the proposed approach and to further our understanding of the computing capabilities of the nervous system. Unlike the conventional synapse, represented as a static entity with a fixed weight, employed in conventional and SRM-based neural networks, a DS is weightless and its strength changes upon the arrival of incoming input spikes. Therefore its efficacy depends on the temporal structure of the impinging spike trains. In the proposed approach, the training of the network free parameters is achieved using an evolutionary strategy where, instead of binary encoding, real values are used to encode the static and DS parameters which underlie the learning process. The results show that spiking neural networks based on both types of synapse are capable of learning non-linearly separable data by means of spatio-temporal encoding. Furthermore, a comparison of the obtained performance with classical neural networks (multi-layer perceptrons) is presented.

Evolutionary Multi-objective Optimization of Spiking Neural Networks

“…The minimal classification error we obtained on the test data is 1.2%, which is a little better than those reported in [5] and [16], where the test error is 1.8% and 2.4%, respectively. Our results are also comparable to those of analog neural networks with two hidden layers [22], where the mean classification error on the test data is 1.15% and 2.87%, respectively, when direct connections between input and output nodes are present or absent.…”

“…In [13], an adaptive GA is adopted to evolve the weights of the SRM model for robot navigation, while in [14], a parallel deferential evolution has been employed to evolve the weights of the SRM. Both weights and delays of an integrate-and-fire (IAF) model with dynamic synapses [15] and a SRM are evolved using an evolution strategy [16]. It is found that the performance of the SNNs are comparable to that of the analog feedforward neural network on two benchmark problems.…”

Evolving Spiking Neural Networks and Neurogenetic Systems for Spatio- and Spectro-Temporal Data Modelling and Pattern Recognition

“…Models of single neurons as well as computational SNN models, along with their respective applications, have been already developed [33,68,73,7,8,12], including evolving connectionist systems and evolving spiking neural networks (eSNN) in particular, where an SNN learns data incrementally by one-pass propagation of the data via creating and merging spiking neurons [61,115]. In [115] an eSNN is designed to capture features and to aggregate them into audio and visual perceptions for the purpose of person authentification.…”

Evolving spiking neural networks for spatio-and spectro-temporal pattern recognition

“…-Moving object recognition [26]; -EEG and fMRI data modelling and pattern recognition [27,47] (fig.8); -Robot control through EEG signals (www.kedri.info, fig.9) and robot navigation [2]; -Sign language gesture recognition (e.g. the Brazilian sign language [ ]; -Risk of event evaluation, e.g.…”