Sistem Pengenalan Pola Angka Menggunakan Metode Delta rule

Abstract

Delta rule is one of the methods in the study of artificial neural networks (ANN). Delta rule has advantages where there is error checking or delta when the learning process is done. This research aims to provide a simple description to young researchers about how to build intelligent systems through one branch of knowledge, namely ANN with the delta rule method.

In this research, delta rule is used as a method that helps computers in the learning process to recognize each target output (target output) of several neurons that carry certain information. The expected output target is a pattern of numbers 1 and 0 using three neurons, where each neuron has a weight value. The ANN weight value is used to store certain information.

The final weight value obtained in this research w₁ is -6.7, w₂ is 1.4, and w₃ is 4.1. With the final weight the system can recognize patterns of numbers 1 and 0 with characteristics that are made diverse as in the truth table that was made in this research.