Monday, November 20, 2017
   
Text Size
Login

Journal of Information and Communication Technology (JICT) Vol. 3, No. 1 June 2004

ADAPTIVE NEURAL NETWORK CLASSIFIER FOR EXTRACTED INVARIANTS OF HANDWRITTEN DIGITS
L. H. Keng and S. M. Shamsuddin
Faculty of Computer Science and Information System
Universiti Teknologi Malaysia, 81310 Skudai, Johor
Tel: 07-5532321 , Fax: 07-5 565044, E-mail: mariyam@frksm.utm.my
E-mail: kelvinlhk@hotmail.com
 
 
ABSTRACT FULL TEXT
We propose an adaptive activation function of neural network classifier for isolated handwritten digits that undergo basic transformations. The utilized network is a backpropagation network with sigmoid and arctangent activation functions. The performance of network with both activation functions is compared. The results show that the network applying an adaptive activation function between layers converged much faster compared to non-adaptive activation functions with 50% iterations reduction. In this study, we also present experimental results of feature extraction between Zernike and d-geometric for better feature representations. Results show that Zernike features are better at representing isolated handwritten digits compared to d-geometric features with an accuracy up to 87%.
 
Keywords: Handwritten Digit, Zernike moments, adaptive activation function
 

 
A SIMULATION STUDY OF DATA DISCOVERY MECHANISM FOR SCIENTIFIC DATA GRID ENVIRONMENT
* A. Abdullah, * M. Othman, * M. N. Sulaiman, * H. Ibrahim, ** and A. T. Othman
*Faculty of Computer Science and Information Technology,
Universiti Putra Malaysia, 43400 Serdang, Selangor
**Faculty of Information Technology and Communication,
Universiti Pendidikan Sultan Idris, 35900 Tanjong Malim, Perak
Email: { azizol, mothman, nasir, hamidah }@fsktm.upm.edu.my
Email :abutalib@upsi.edu.my
 
ABSTRACT FULL TEXT
Research in the area of grid computing has given us various ideas and solutions to address the requirements in a modern scientific computing community that managed massive amounts of a very large data collections in a geographically distributed environment. Data Grids mostly deal with large computational problems and provide geographically distributed resources for large-scale data-intensive applications that generate large data sets. A number of research groups are working on the data distribution problems in Data Grids and they are investigating the data replication approaches on the data distribution. This leads to a new problem in discovery and access to data in Data Grid environment. To address this problem we have developed a model to study various discovery mechanisms and investigate these mechanisms for Dynamic Scientific Data Grid Environments using our Grid Simulator. In this paper, we illustrate our model and our Grid Simulator.
 
Keywords: replication, data grids, data discovery
 

 
ON NETWORK FLOW PROBLEMS WITH CONVEX COST
V. A. Nguyen and Y. P. Tan
School of Electrical & Electronic Engineering
Nanyang Technological University, Singapore
Nanyang Technological University
Block S1 , Nanyang Avenue, Singapore 639798
 
ABSTRACT FULL TEXT
Minimum cost flow (MCF) problem is a typical example of network flow problems, for which an additional constraint of cost is added to each flow. Conventional MCF problems consider the cost constraints as linear functions of flow. In this paper, we extend the MCF problem to cover cost functions as strictly convex and differentiable, and refer to the problem as convex cost flow problem. To address this problem, we derive the optimality conditions for minimising convex and differentiable cost functions, and devise an algorithm based on the primal-dual algorithm commonly used in linear programming. The proposed algorithm minimises the total cost of flow by incrementing the network flow along augmenting paths of minimum cost. Simulation results are provided to demonstrate the efficacy of the proposed algorithm.
 
Keywords: Minimum cost flow problem, convex cost functions, primal-dual, algorithm, network flow.
 

 
PARALELL R-POINT IMPLICIT BLOCK METHOD FOR SOLVING HIGHER ORDER ORDINARY DIFFERENTIAL EQUATIONS DIRECTLY
*Z. Omar and **M. Sulaiman
* Faculty of Quantitative Sciences,
Universiti Utara Malaysia, 06010 UUM Sintok, Kedah
** Department of Mathematics, Universiti Putra Malaysia,
43400 UPM Serdang, Selangor
Tel: 04-928 4900, Fax: 04-928 5779, Email: zurni@uum.edu.my
 
ABSTRACT FULL TEXT
Most of the existing methods for solving ordinary differential equations(ODEs) of higher order are sequential in nature. These methods approximate numerical solution at one point at a time and therefore do not fully exploit the capability of parallel computers. Hence, the development of parallel algorithms to suit these machines becomes essential. In this paper, a new method called parallel R-point implicit block method for solving higher order ODEs directly using constant step size is developed. This method calculates the numerical solution at more than one point simultaneously and is parallel in nature, thus suitable for parallel computation. Computational advantages are presented comparing the results obtained by the new method with that of conventional 1-point method. The numerical results show that the new method reduces the total number of steps and execution time. The accuracy of the parallel block and the conventional 1-point methods are comparable particularly when finer step sizes are used.
 
Keywords: parallel R-point, implicit block method, higher order ODEs, directly.
 

 
RICE YIELD CLASSIFICATION USING BACKPROPAGATION NETWORK
*P. Saadª, ***N. K. Jamaludin, **S. S. Kamarudin, ***A. Bakri, and *N. Rusli
*Northern Malaysia University College of Engineering(KUKUM)
Taman JKKK, Kubang Gajah, 02600 Arau, Perlis, Malaysia.
**Faculty of Information Technology
Universiti Utara Malaysia, Sintok, Kedah, Malaysia
***Faculty of Computer Science and Information System
University Technology Malaysia, Skudai, Johor, Malaysia
puteh@kukum.edu.my, khairah jamal@yahoo.com
sakira@uum.edu.my, aryati@fsksm.utm,my
nursalasawati@kukum.edu.my
 
ABSTRACT FULL TEXT
Among factors that affect rice yield are diseases, pests and weeds. It is intractable to model the correlation between plant diseases, pests and weeds on the amount of rice yield statistically and mathematically. In this study, a backpropagation network (BPN) is developed to classify rice yield based on the aforementioned factors in MUDA irrigation area Malaysia. The result of this study shows that BPN is able to classify the rice yield to a deviation of 0.03.
 
Keywords: Backpropagation Network, Classification, rice yield, pests, diseases, and weeds.
 

 
SIMULATION-AIDED PRODUCTION OF AIRCRAFTS COMPOSITE PARTS
*M. K. Mohd-Nawawi, **R. M. Tahar, and ***Z. Omar
Faculty of Quantitative Science
Universiti Utara Malaysia, 06010 Sintok, Kedah, Malaysia.
*Tel: 04-9286938, Fax: 04-9283705
**Tel: 04-9285083, Fax: 04-9283705, Email: ratman@uum.edu.my
***Tel: 04-9285124/4900, Fax: 04-9283705, Email:
zurni@uum.edu.my
 
ABSTRACT FULL TEXT
Simulation has been successfully applied in many different areas such as manufacturing, service system, healthcare, transportation, supply chain, and etc. The opportunities to cut costs and to improve service levels in these sectors are tremendous by applying this technology. One of the largest application areas for simulation modelling is that of manufacturing systems, which has been applied for more than 40 years. This study is conducted at the local manufacturing plant that manufactures composite products for the aerospace industry. In this study, simulation technique is applied to model the existing system to simulate the design and operational policies of the plant production process, which can be used to improve the performance of the different activities at the plant. The model focused on activities at Secondary Manufacturing Area (SMA) of aircrafts composites parts.
 
Keywords: Simulation, performance, plant, production, aircraft.

Universiti Utara Malaysia Press
 Universiti Utara Malaysia, 06010 UUM Sintok
Kedah Darul Aman, MALAYSIA
Phone: +604-928 4816, Fax : +604-928 4792

All Right Reserved. Copyright © 2010, Universiti Utara Malaysia Press