Enhanced conjugate gradient methods for training MLP-networks

Izzeldin, H. and Asirvadam , Vijanth Sagayan and Saad , Nordin (2010) Enhanced conjugate gradient methods for training MLP-networks. In: Research and Development (SCOReD), 2010 IEEE Student Conference on, 13-14 December 2010, Putrajaya.

[thumbnail of Published IEEE paper] PDF (Published IEEE paper)
05703989.pdf - Published Version
Restricted to Registered users only

Download (721kB)
Official URL: http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?arn...


The paper investigates the enhancement in various conjugate gradient training algorithms applied to a multilayer perceptron (MLP) neural network architecture. The paper investigates seven different conjugate gradient algorithms proposed by different researchers from 1952-2005, the classical batch back propagation, full-memory and memory-less BFGS (Broyden, Fletcher, Goldfarb and Shanno) algorithms. These algorithms are tested in predicting fluid height in two different control tank benchmark problems. Simulations results show that Full-Memory BFGS has overall better performance or less prediction error however it has higher memory usage and longer computational time conjugate gradients.

Item Type: Conference or Workshop Item (Paper)
Uncontrolled Keywords: BFGS;Broyden Fletcher Goldfarb and Shanno;MLP;MLP networks;conjugate gradient methods enhancement;fluid height prediction;gradient training algorithms;multilayer perceptron;neural network architecture;tank benchmark problems;gradient methods;learning (artificial intelligence);multilayer perceptrons;
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering
Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Departments / MOR / COE: Centre of Excellence > Centre for Automotive Research
Departments > Electrical & Electronic Engineering
Depositing User: Dr Vijanth Sagayan Asirvadam
Date Deposited: 05 Dec 2011 03:01
Last Modified: 19 Jan 2017 08:23
URI: http://scholar.utp.edu.my/id/eprint/4632

Actions (login required)

View Item
View Item