Agent based Modeling for Activation Function in Enhancement Logic Programming in Hopfield Neural Network

Abstract

Artificial Neural Network (ANN) uses many activation functions to update the state on neuron. The research and engineering have been used activation functions in the artificial neural network as the transfer functions. The most common reasons for using this transfer function were its unit interval boundaries, the functions and quick computability of its derivative, and several useful mathematical properties in the approximation of theory realm. Aim of this study is to figure out the best robust activation functions to accelerate HornSAT logic in the Hopfield Neural Network's context. In this paper we had developed Agent-based Modelling (ABM) assessed the performance of the Zeng Martinez Activation Function (ZMAF) and the Hyperbolic Tangent Activation Function (HTAF) beside the Wan Abdullah method to do Logic Programming (LP) in Hopfield Neural Network (HNN). These assessments are carried out on the basis of hamming distance (HD), the global minima ratio (zM), and CPU time. NETLOGO 5.3.1 software has been used for developing Agent-based Modeling (ABM) to test the proposed comparison of the efficaecy of these two activation functions HTAF and ZMAF.