Journal of Digital Information Management


Vol No. 20 ,Issue No. 4 2022

Utilization of SAM-based Network for Developing Function Approximation
Minoru Motoki, Hirohito Shintani, Kazunori Matsuo, Thomas Martin McGinnity
Deparment of Information Communication and Electronc Engineering National Institute of Technology Kumamoto College, Kumamoto, Japan., Dept. of Control and Information Systems Engineering National Institute of Technology Kumamoto College, Kumamoto, Japan.
Abstract: We have previously reported progress in developing a multilayer SAM spiking neural network and a training algorithm, suitable for implementation on an FPGA with “On- Chip Learning”. Here we report on utilization of a SAM -based network for continuous function approximation, which to date has proved difficult to achieve on a LIF type spiking neural network, by using a spike coding approach called ‘NFR-coding’. We demonstrate “interpolated XOR” and 3-polynominal function approximation of this SAM network in computational experiments. It is demonstrated that the SAM network has the capability to perform these function approximations to high accuracy.
Keywords: Spiking neural network(SNN), SAM Neuron Model, FPGA implementation, On-chip learning, Function Approximation Utilization of SAM-based Network for Developing Function Approximation
DOI:https://doi.org/10.6025/jdim/2022/20/4/148-155
Full_Text   PDF 3.26 MB   Download:   148  times
References:

[1] Shigematsu, Y. & Matsumoto, G. (2000) Temporal learning rule and dynamicneural network model. Applied Mathematics and Computation, 111, 209–218.

[2] Motoki, M., Soh, S.J.H., Matsuo, K., Shintani, H. & McGinnity, T.M. (Under review) A supervised training algorithm for multilayer SAM spiking neural network. IEEE Transactions on Neural Networks and Learning Systems.

[3] Bohte, S.M., Kok, J.N. & La Poutré, H.L. (2002) Error-backpropagation in temporally encoded networks of spiking neurons. Neurocomputing, 48, 17–37.

[4] Iannella, N. & Back, A.D. (2001) [Special Issue] A spiking neural network architecture for nonlinear function approximation. Neural Networks. Neural Network, 14, 933–939.

[5] Motoki, M. & Matsuo, K. (2014) (in Japanese) An error-correction learning algorithm for SAM neuron model and its HDL design Workshop on Circuits and Systems, Vol. 27, pp. 244–249.

[6] Maass, W. & Bishop, C.M. (2001). Pulsed Neural Networks. MIT Press: Cambridge, USA.