| Font + -
You are here:   HomeProgrammes OfferedPostgraduateMSc. (IT Technopreneurship)Research

Research

Category: Research

Research Review: Analysis Of Biomarker Identification Through Shrinkage Estimation And deABC (Artificial Bee Colony With Differential Evolution) From Mass Spectrometry
2014-04-25
Time 1000 until 1130
Meeting Room 7th Floor
Syarifah Adilah Mohamed Yusoff
Prof. Dr. Rosni Abdullah
Dr. Ibrahim Venkat
Mass spectrometry technique is gradually gaining momentum among the recent techniques deployed by several analytical research labs, especially to study biological or chemical properties of complex structures such as protein sequences. This advancement also embarks the discovery of biomarkers through accessible body fluid such as serum, saliva and urine. Due to high-throughput data, a given molecule species can give rise to a series of inter-related peaks in a mass spectrum. A sample data can result in a very complex spectrum with many interrelated and overlapping peaks. This scenario will degrade the potential features to be extracted from the spectrum. The spectrum also suffers with high dimensionality data relative to small samples size. Constructing a good classification model to predict biomarkers from disease and normal cases requires well-discriminated and independent potential features. This study focused on two stages of mass spectrometry pipelines. Firstly on the feature extraction where several peaks across different samples will be assembled and calibrated based on their strong correlation coefficient and at the same time reduce dimensionality of data. Secondly on the feature selection through wrapper techniques to search for parsimonious features through a learning model. Shrinkage estimation of covariance was proposed to evaluate the discriminant characteristics among peaks of mass spectrometry data for feature extraction. Meanwhile for feature selection, a sophisticated computational technique that mimics survival and natural processing known as Artificial Bee Colony (ABC) and Differential Evolution (DE) were hybridised and further integrated with linear SVM classifier for this biomarker discovery analysis. The proposed method is tested with real-world mass spectrometry datasets such as ovarian, liver and TOX dataset to evaluate the discrimination power, accuracy, sensitivity and also specificity.
N/A
none
2013/14
2
N/A
Read more...
 
Category: Research

Research Review: An Intelligent Framework for Enhancing The Video Transmission Over Vehicular Ad Hoc Networks: Multipath and Optimization
2014-04-02
Time 1430 until 1600
Meeting Room 7th Floor
Walid Shaher M. Yousef
Assoc. Prof. Muhammad Rafie Hj. Mohd. Arshad
N/A
In recent decades, the field of wireless communication has been a pointed increase in both industrial research and commercial applications. Progress in this area has significantly changed the daily life of people around the world. In this context, Vehicular Ad Hoc Networks (VANETs) where vehicles are able to communicate with each others, emerge as a promising wireless technology able to enhance the vision of drivers and offer a larger telemetric horizon. There are several methods to enhance performance in video streaming transmission via ad hoc networks such as multipath schemes which could be an appropriate solution. In order to increase performance from several sides, multipath is used to deliver video content at the same time. VANET is one of most interested in wireless networks, where Multipath routing for video transmission over VANET discussed by researchers as critical gaps are still found in the recent literature. For this reason, this study primarily focuses on video transmission over VANET using intelligent methods. Moreover, the preliminary results indicate that the intelligent solution in VANET has a reasonable capability to address and fill the gaps in the recent literature. Additionally, OMNeT++ and MATLAB are used to simulate the proposed solutions in order to evaluate the solution.
N/A
none
2013/14
2
N/A
Read more...
 
Category: Research

Proposal Review: Two Dimensional Generative Models based on Membrane Computing
2014-04-09
Time 1100 until 1230
Meeting Room 7th Floor
Pradeep Isawasan
Dr. Ibrahim Venkat
Dr. Nurul Hashimah Ahamed Hassain Malim
The field of Membrane Computing was initiated by Gheorge P ? un around the year 2000. The computing model in this field is called a P system in honour of its originator. A P system is a computing device which consists of several cell-like membranes placed inside a skin membrane with a hierarchical arrangement and with objects placed in the regions delimited by the membranes. In the basic model, objects are allowed to evolve by evolution rules and can communicate from one region to another, thus leading to a computation of an output. P systems have proved to be a rich theoretical framework to study many computational problems besides giving a new impetus to formal language theory. In studies on image analysis, syntactic models of two dimensional array generation have occupied an important place in an area where techniques or methods on information extraction from two dimensional arrays is investigated. In 2003, Ceterchi et al. provided a link between two dimensional grammars of formal language theory and P systems by extending the string- objects P systems to array-objects P systems, called array rewriting P systems. Motivated by this model, here in this thesis we investigate the generative capacity of the array rewriting P systems by endowing them with additional features from formal language theory such as parallel rewriting and permitting features. We also consider applying the permitting features in grammars called context-free triangle-tiled two dimensional array grammars, which are introduced by Subramanian et al. (2012) for the problem of generation of two dimensional triangletiled arrays. As an application, we consider the problem of image segmentation which is the process of assigning a label to every pixel in an image such that pixels with the same label share certain visual characteristics. In the context of formal language theory, the same process is viewed where the label is the symbol for an object (pixel) in a two dimensional plane (digital image). In this thesis we apply a variant of P systems called tissue-like P systems introduced by MartínVide et al. (2003) for the region-based segmentation problem in two dimensional hexagonal arrays. Thus this thesis is dedicated to the study of the generative capacity of the variants of array P system models and its application to image analysis.
N/A
none
2013/14
2
N/A
 
Category: Research

Research Review: The Development of New Imperceptible Image-based Steganographic Algorithms
2014-03-18
Time 1100 until 1230
Meeting Room 7th Floor
Samer Hassan Suleiman Atawneh
Assoc. Prof. Dr. Putra Sumari
N/A
Digital image-based steganography is the practice of hiding secret information into digital images with the intention of communicating hidden information. It has become significant in today’s digital world where information is frequently and easily exchanged through the Internet, email, and other computer usage. These electronic communication means, which are susceptible to attacks and eavesdropping, make current security measures more important than ever before, where security problems, such as modification and forgery, have reached critical extents. The need for creating effective methods for image-based secret sharing led to the new incentive research in the area of image steganography. This study investigates the current state-of-the-art image-based steganographic schemes to highlight the key concepts behind them and provide the advantages and drawbacks of the different steganography algorithms. The main goal of this research is to enhance the imperceptibility and increase the embedding payload by providing new secure and efficient algorithms that struggle to reduce the drawbacks of current image-based steganographic schemes and narrow down the margin of embedding payload and distortion. The first proposed algorithm exploits the complexities between pixel pairs of the cover image to embed different-bases secret digits. In this algorithm, one pixel at most of each pixel pair is modified to hold the secret digit and hence introduces less-distorted image. The second proposed algorithm exploits the frequency-domain adjacent coefficients of the cover image to embed base-5 secret digits by changing the value of at most one coefficient of each coefficient pair to embed the secret digit. The third proposed algorithm uses the spatial and frequency domains to develop a new hybrid information hiding algorithm for digital images where the frequency-coefficients are utilized to embed secret information. In this hybrid algorithm, a chaotic map is utilized to scramble the secret information before the embedding procedure takes place. The proposed algorithms are tested against relevant performance metrics. The results are promising and show that the developed algorithms have improved the imperceptibility and embedding payload as compared to other similar existing techniques.
N/A
none
2013/14
2
N/A
Read more...
 

Related Topics

Contact Us

Home
Address:
School of Computer Sciences
Universiti Sains Malaysia
11800 USM, Penang, Malaysia
Tel: (+604) 653 3888 ext. 3647/3610
Fax: (+604) 657 3335

Best View

Screen Resolution:
1280 x 1024 or above with

Firefox Logo
FireFox6
or above

Google Chrome Logo
Google
Chrome

Internet Explorer 7 Logo
IE 6
or above

Opera Logo
Opera11.5
or above

**For IE users, please UNCHECK the `Compatibility View` setting at Tools menu for a proper view.

Login