04-801-J4   Artificial Neural Networks

Course description

The topics of deep learning and artificial intelligence have become very popular due to their potential applications in many practical situations. The reason for the surge in these areas is the availability of huge computing power and large volumes of data. The key architectures used in these applications are deep neural networks and recurrent neural networks. The objective of this course is to trace the historical developments of artificial intelligence leading to artificial neural networks (ANN). The course introduces the basic concepts and models of ANN for solving simple pattern recognition problems. In particular, it includes analysis of feedforward and feedback neural networks, involving the key concepts of backpropagation learning and Boltzmann machine, and the pattern recognition tasks they perform. The course concludes with a discussion on the evolution of ANN architectures from learning to deep learning.

Learning objectives

This course deals with the historical developments of artificial intelligence leading to artificial neural networks (ANN) and introduces the basic concepts and models of ANN for solving simple pattern recognition problems.

Content details

• Background to ANN and Parallel and Distributed Processing (PDP) models; Basics of ANN including terminology, topology, and learning laws (3 lectures)
• Analysis of Feedforward Neural Networks (FFNN) including linear associative networks, perceptron networks, multilayer perceptron, gradient descent methods, and backpropagation learning (5 lectures)
• Analysis of Feedback Neural Networks (FBNN) including Hopfield model, state transition diagram, stochastic networks, Boltzmann learning law (5 lectures)
• Evolution of ANN architectures - from learning to deep learning (1 lecture)

Prerequisites

The course is mostly self-contained, but some background in linear algebra and probability theory will be useful. Basic mathematics covered in engineering programs is assumed.

Faculty

B. Yegnanarayana