Please use this identifier to cite or link to this item:
Title: Artifical Neural Network Application for Printed Character Recognition Using Back-Propagation Algorithm
Authors: Ohri, Anil
Supervisor: Radhakrishna, M
Soni, M.K.
Keywords: Neural computing;Artificial neural network;Computational architecture;Computer Science
Issue Date: 11-Sep-2007
Abstract: The advent of neural computing has opened up new frontiers for research; in the field of A.I. Artificial neural network (ANN) is a unique computational architecture. These have been applied to solve the problems which are computationally very difficult with conventional methods and which are non-algorithmic, fuzzy in nature and no clear cut mathematical model.The main objective of neural modeling was to produce computational systems that performs human brain like function efficiently. Many powerful learning techniques such as Back propagation algorithm, delta rule etc. are available to generate the neural networks from training examples. Back propagation (BP) is one of the most commonly used systematic method to train multilayered neural network. BP algorithm has limitation that it fail to train a network with non-differentiable activation function.In this dissertation work B.P. algorithm has been used for multiplayer feed forward network training to solve printed character recognition problems. The different training pairs for different printed characters have been taken to train the ANN using B.P. algorithm and comprehensive performance evaluation of the limitation and capabilities of multiplayer feed forward networks for printed character recognition has been done.
Appears in Collections:Masters Theses@CSED

Files in This Item:
File Description SizeFormat 
m91156.pdf7.75 MBAdobe PDFView/Open    Request a copy

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.