A Comprehensive Performance Analysis on Artificial Neural Networks

Abstract

Neural Networks have dominated the sphere of Machine Learning and computerized decision trends for the past decade. The most straightforward neural architecture holds the key to some of humanity's most complex and vexing problems. When this concept of mimicking the human brain in digital or machine interpretation was first materialized in the late 1940s, the analysts were crippled by the technological reach of their time. But slowly, the advent of faster computational prowess and memory extensions paved the way for the intuitive backpropagation process in 1975, which was the first robust training procedure globally accepted. It becomes the fundamental requisite of almost all technological interactions we experience every day. Understanding the reflective activities, of an Artificial Neural Network is the first step toward more profound innovations and discoveries in machine learning. This paper specifically attempts to give an insight on various types of Neural Networks. Pros and cons of each Neural Network is summarized including their performance analysis in several application areas.