Universiti Teknologi Malaysia Institutional Repository

The effect of weights initialization on osteo arthritis classification

Shamsuddin, Siti Mariyam and Ahmed, Falah. Y. H. (2010) The effect of weights initialization on osteo arthritis classification. In: Asia Modelling Symposium 2010 : 4th Asia International Conference on Mathematical Modelling and Computer Simulation, 26-28 Mac 2010, Borneo Island, Kota Kinabalu, Sabah.

Full text not available from this repository.

Official URL: http://gmm.fsksm.utm.my/~mariyam/PUBLISHED_PAPERS_...

Abstract

Standard Back propagation Algorithm (BP) is a widely used in Multilayer Perceptron by the practitioners despite its existence for almost four decades. It is proven to be very successful in diverse applications, such as Osteoarthritis diagnoses. Osteoarthritis diagnoses are one of the most frequent causes of physical disability among adults. Therefore, this study proposes Osteoarthritis diagnoses classification with improved structures of BP network by proposing acceleration parameters using adaptive learning. The proposed adaptive learning involves two mechanisms: weights initialization and the usage of logarithm activation function to reduce the error rate and convergence time. From the experiments, we found that by selecting appropriate initial weights can lead to feasible results and faster learning for Osteoarthritis diagnoses classification. These are proven by the experiments conducted on the enhanced BP, which is better than a standard BP in terms of faster convergence and less errors generated.

Item Type:Conference or Workshop Item (Paper)
Uncontrolled Keywords:artificial neural network, backpropagation algorithm, classification, weights initialization, osteoarthritis
Subjects:Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Divisions:Computer Science and Information System (Formerly known)
ID Code:24355
Deposited By: Liza Porijo
Deposited On:26 Sep 2012 02:08
Last Modified:26 Sep 2012 02:08

Repository Staff Only: item control page