Evolutionary approach in Assembler by Implementation of Neural Networks

Abstract

A neural network is represented using the Assembler Encoding approach as a straightforward computer programme called the Assembler Encoding Programme. The goal of this programme is to create a Network Definition Matrix, which has all the data required to build a neural network. In order to create the programmes and subsequently the neural networks, evolutionary techniques are used.Finding the ideal number of neurons for a neural network is one of the difficulties in Assembler Encoding. The current implementation of Assembler Encoding relies on an inefficient and time-consuming way to deal with this issue. The report offers four other approaches to address this problem, though. Experiments were performed utilising a predator-prey problem to assess various solutions. The network's design, including the number of layers, the number of neurons in each layer, and the connections between neurons, is specified by the Assembler Encoding Programme. These rules are expressed in a low-level language that resembles machine code very closely.The Network Definition Matrix is produced by running the Assembler Encoding Programme, and it acts as a design for the neural network. Typically, the matrix contains data on the biases and weights of the connections between neurons.Thepaper track the Assembler Encoding Programmes are frequently improved and evolved using evolutionary techniques like genetic algorithms. The programmes are iteratively enhanced through evolutionary processes to create neural networks that display desirable behaviour for certain tasks or issues.