Please use this identifier to cite or link to this item:
Title: Implementation of Hebbian - LMS Algorithm
Authors: Vartika
Supervisor: Sakshi
Keywords: Machine Learning;Neural Networks;Supervised, Unsupervised Learning;LMS;Hebbian-LMS
Issue Date: 22-Sep-2017
Abstract: In most of the engineered network systems, a set of operations interact with each other in complex manners that can contain multiple types of relationships, depending on time and include other types of complexities. Such network systems comprise multiple subsystems and multiple layers of connectivity, and are generally called as multi-layer networks. There exist multi-layer neural networks which are feed forward artificial neural network model that maps sets of input data onto a set of appropriate outputs. it has multiple layers with each layer fully connected to the next one. An artificial multi layers neural network contains three layers broadly defined as input layer, hidden layer and output layer in the same order as written. These network systems can be trained using both supervised and unsupervised algorithms. The widely used supervised learning in these systems is LMS learning algorithm whereas the unsupervised learning used is Hebbian learning. A form of LMS algorithm can be established to achieve unsupervised learning. In this way LMS can be used to implement Hebbian learning. This implementation of combined algorithms is called as Hebbian-LMS learning algorithm. Combining the two algorithms creates a new unsupervised learning algorithm that has application in practical engineering problems. In this thesis, an artificial neural network is considered, whose hidden layer weights are adjusted using Hebbian-LMS algorithm and the output layer is trained using the original supervised LMS algorithm and the results are recorded for various datasets using this approach as training algorithm in order to determine the feasibility of proposed algorithm in networks required to solve some practical engineering problems.
Appears in Collections:Masters Theses@ECED

Files in This Item:
File Description SizeFormat 
4892.pdf1.04 MBAdobe PDFThumbnail

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