Title: Weightless for neural networks based on multi-valued probabilistic logic for node for pattern recognition

Authors: Nadia Nedjah; Luiza de Macedo Mourelle; Tarso Mesquita Machado

Addresses: Department of Electronics Engineering and Telecommunications, State University of Rio de Janeiro, Rio de Janeiro, Brazil ' Department of Systems Engineering and Computation, State University of Rio de Janeiro, Rio de Janeiro, Brazil ' Post-Graduate Program of Electronics Engineering, State University of Rio de Janeiro, Rio de Janeiro, Brazil

Abstract: Weightless neural networks draw inspiration from the dynamic and adaptable nature of synaptic connections observed in biological neural systems. By reducing dependence on fixed weights, these kinds of networks allow enhancing computational efficiency. Yet, their unconventional architecture poses challenges in developing effective architecture and training algorithms and theoretical frameworks. There are many such architecture. Here, we concentrate on those based on multi-valued probabilistic logic neurons (MPLN). This work studies the impact of the hyper-parameters of this kind of network in order to design an efficient and concise network topology. The present work further proposes a modification in the MPLN network for multi-class problems, termed the Mod-MPLN network. The Mod-MPLN network is defined by a change in the network training algorithm and by the inclusion of a specific discriminator at the network output. The performance is evaluated based on the implementation of several MPLN architectures for two applications. The performance comparison proves that the Mod-MPLN outperforms the WiSARD weightless neural network in terms of classification accuracy. For the multi-class application and considering the entire dataset, Mod-MPLN achieved an average accuracy of 97.84% against 68.58% for MPLN and 85.71% for WiSARD.

Keywords: weightless neural network; probabilistic logic node; pattern recognition.

DOI: 10.1504/IJBIC.2024.141461

International Journal of Bio-Inspired Computation, 2024 Vol.24 No.2, pp.80 - 97

Received: 27 Feb 2024
Accepted: 27 May 2024
Published online: 13 Sep 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article