Pusat Pengajian Sains Komputer - Tesis

Browse

Now showing 1 - 20 of 72

Compute Language Interface: A Transparent Wrapper Library For Multi Cpu-Gpu
(2013-03)
Ooi, Keng Siang
The Graphics Processing Unit (GPU) processing capability is getting more powerful than before. Compute intensive and data parallelism applications are proven to perform better on the GPU than on the Central Processing Unit (CPU). However, available General-Purpose Computing on Graphics Processing Unit (GPGPU) programming frameworks which are available publicly are unable to reach beyond the single computer limitation to utilize multiple CPUs and GPUs at different computers in a distributed computing system easily. This study presents the Compute Language Interface (CLI) which is a wrapper library that enables the existing OpenCL applications access to all available CPUs and GPUs in a distributed computing system through Message Passing Interface (MPI) transparently. It is designed to improve the scalability of the OpenCL applications on a distributed computing system while maintaining the same set of application programming interface (API) in the original OpenCL library. The applications can access all available CPUs and GPUs in different computers in a distributed computing system as ifall the CPUs and GPUs are in the same computer.
Social Sciences, Arts And Humanities Research Collaboration: The Malaysian Coverage
(2014)
Hussain, Husriati
As a developing country, Malaysia emphasizes the need for university researchers to be more socially oriented and highly encourages them to seek collaboration with other organizations. Collaboration is used as the main requirement and indicator in almost Malaysian grant scheme and Malaysian research assessment, such as Higher Institutions Centre of Excellence (HICoE) Grant, Long Term Research Grant Scheme (LRGS), Malaysian Research Assessment (MyRA), National Higher Education Strategic Plan, etc. The main reasons behind this are because collaboration may contribute to creative, wealth and innovative ideas, unique experiences, and great exposures, which consequently will give a high impact on the nation's research productivity and development. In this regard, this study employed bibliometric methods of co-authorship analysis to measure the patterns of collaboration in Social Sciences, Arts and Humanities in Malaysia from 2007 to 2011. Based on the 2,280 publications, patterns of collaboration were analyzed with respect to 7 relevant areas: co-authorship patterns, national and international collaboration, country collaboration, institutional collaboration, sectoral collaboration, intra-disciplinary and interdisciplinary collaboration, and extent of collaboration. The results may help the government and universities to plan an appropriate Malaysia Research Policy and manage the R&D funding and resources effectively and efficiently. The
A Framework For Automatic Code Switching Speech Recognition With Multilingual Acoustic And Pronunciation Models Adaptation
(2014-05)
Ahmed, Basem H. A.
Recognition of code-switching speech is a challenging problem because of three issues. Code-switching is not a simple mixing of two languages, but each has its own phonological, lexical, and grammatical variations. Second, code-switching resources, such as speech and text corpora, are limited and difficult to collect. Therefore, creating codeswitching speech recognition models may require a different strategy from that typically used for monolingual automatic speech recognition (ASR). Third, a segment of language switching in an utterance can be as short as a word or as long as an utterance itself. This variation may make language identification difficult. In this thesis, we propose a novel approach to achieve automatic recognition of code-switching speech. The proposed method consists of two phases, namely, ASR and rescoring. The framework uses parallel automatic speech recognizers for speech recognition. We also put forward the usage of an acoustic model adaptation approach known as hybrid approach of interpolation and merging to cross-adapt acoustic models of different languages to recognize code-switching speech better. In pronunciation modeling, we propose an approach to model the pronunciation of non-native accented speech for an ASR system. Our approach is tested on two code-switching corpora: Malay-English and Mandarin-English.
Acltshe-Amts: A New Adaptive Brain Tumour Enhancement And Segmentation Approaches
(2014-05)
Alkhafaji, Ali Fawzi Mohammed Ali
Brain tumor subregion segmentation from multimodal Magnetic Resonance (MR) images is of great interest for better tumor diagnosis. Multilevel thresholding is one of the prominent approaches used for brain image segmentation. Currently, when applying multilevel thresholding for brain tumor segmentation, two important problems must be carefully addressed. First, the MR brain images suffer from sensitivity to intensity inhomogeneity, poor contrast, and hidden details, which corrupt the original MR image during capturing. Second, the conventional multilevel thresholding approaches, including optimization-based thresholding approaches, have several main issues, such as manual adjustment of multilevel thresholds, dedicating single thresholding criteria as objective functions, leading to the bias of the thresholding towards a specific type of MR image, and the requirement to tune a large number of control parameters. In this thesis, a two-stage approach is proposed to address these issues. In the first stage, a new image enhancement approach called Adaptive Clip Limit Tile Size Histogram Equalization (ACLTSHE) is proposed to improve contrast, highlight the hidden details, and achieve homogenized intensity distribution of MR images. The ACLTSHE integrates Contrast-Limited Adaptive Histogram Equalization, Multi-Objective Whale Optimization Algorithm, Discrete Entropy (DE), Peak Signal-to-Noise Ratio (PSNR), and Structure Similarity Index (SSI) to improve the quality of MR images while preserving the original structure of the MR images. In the second stage, a new approach called Adaptive Multilevel Thresholding Segmentation (AMTS) is proposed for unsupervised brain tumor xxi subregion segmentation from normal brain tissue. The AMTS approach segments and extracts the whole tumor, core tumor, and enhanced tumor regions from the brain MR images, integrating the Multi-Objective Grasshopper Optimization algorithm, Kapur Entropy, Cross-Entropy, and Localized active contour.
Classification Of Cpg Island And Promoter Regions Using Rare K-Mer Motifs
(2015-07)
Mohamed Hashim,Ezzeddin Kamil
Empirical analysis on DNA k-mers is proven to be an effective means to discover functional elements in the human genomes. Among the empirical works, “rare k-mer" (RKM) is a very interesting subject to be studied due to their unique sequence properties. RKMs were referred as DNA k-mers (of k=7 to 11) that have a low frequency in k-mer frequency distributions of mammalian genomes, yet there are large variations of their mass at the lower k-mer spectra. Our first objective is to discover RKM motifs in the human genome; to identify potential RKM computational applications in biology; and to infer representations for the identified applications. In short, the first goal was achieved by using comparative strategy and several bioinformatic tools (of UCSC browsers, EpiGRAPH, and Galaxy) which correlated RKMs with several genomic features of CpG Islands (CGIs), promoter, 5’ Un- Translated Regions (5’UTRs), and open chromatin regions; and by using intrinsic string mining approaches which identified several unique RKM topological, compositional, and clustering properties in the correlated genomic features.
Cloud Resource Management Framework Using Monarch Butterfly Harmony Search And Case Based Reasoning
(2017-08)
Ahmed Mohamed Ghetas, Mohamed Rezk
Cloud services have evolved rapidly and some have adopted a multi-tier architecture for flexibility and reusability. Various rule- and model-based approaches have designed to manage quality of service for these services. A few of existing resource management approaches aim to increase the cloud provider (cp) service provisioning profits. However, they are based on local search optimization algorithms, which may not obtain the best resource provisioning decision in a large-scale cloud environment. This research proposes a new resource optimization and provisioning (rop) framework to detect, solve the bottlenecks, and satisfy the service-level qos requirements of several multi-tier cloud services and to increase the cp service provisioning profits. The rop framework consists of two main components: global resource optimizer (gro) and resource identifier (ri). This research enhances the butterfly optimization algorithm and plugs the resulting algorithm into the rop as a gro. In addition, a new ri is developed using case-based reasoning and is then plugged into the rop framework. To demonstrate the effectiveness of the proposed rop against rule- and model-based approaches, a prototype running on a cloud platform is developed, and a workload generator and multi-tier service model are adopted.
Real-Time Capable Multi-Hop Media Access Control Protocol For Smart Home Environment
(2017-11)
M.Shukeri, Nurulfaizal
The Wireless Sensor Network (WSN) is a technology that is now often highlighted and various studies have been done to apply in life such as environmental monitoring, security and military applications. These include the study of the Internet of Things (loT) and Smart Home, where it is now gaining popularity in the research environment. The combination of home appliances such as lights, gates and closed circuit, would be able to make the future home not just smart, but smarter in energy consumption and secure. Nevertheless, to apply WSN in the Smart Home Environment, WSN protocols need to support real-time characteristics. This protocol must also be capable of transmitting time-sensitive data such as audio and video at low bit rates through a multi-hop network where coverage can be expanded.
Hybrid Of Optimized Random Forest And Extreme Gradient Boosting For Online Learning Style Classification
(2019-03)
Haziqah Shamsudin
Educational Data Mining (EDM) have raised a lot of attention among researchers since the last few decades. EDM is used to gain more insight into the behavior of learners by building models based on data collected from learning tools which result in improving learning system to be more personalized and adaptive. Learning style of specific users in the online learning system is determined based on their interaction and behaviour towards the system. Felder-Silverman’s learning style model is the most common online learning theory used in determining the learning style. Initially, in determining the users’ learning styles, users are asked to fill in the questionnaires which is designed to learn their learning style at the end of the learning sessions. However, this method is time consuming and the result are not reliable due to the human factors behavior. Thus, the researchers started to study the learning style by using an automated approach in which the activity log files are collected in order to understand the interactivity behaviour of the users with the system.
Multimodal Sentiment Analysis Of Social Media Through Deep Learning Approach
(2020-06)
An, Jieyu
Multimodal Data, Characterized By Its Inherent Complexity And Heterogeneity, Presents Computational Challenges In Comprehending Social Media Content. Conventional Approaches To Sentiment Analysis Often Rely On Unimodal Pre-Trained Models For Feature Extraction From Each Modality, Neglecting The Intrinsic Connections Of Semantic Information Between Modalities, As They Are Typically Trained On Unimodal Data. Additionally, Existing Multimodal Sentiment Analysis Methods Primarily Focus On Acquiring Image Representations While Disregarding The Rich Semantic Information Contained Within The Images. Furthermore, Current Methods Often Overlook The Significance Of Color Information, Which Provides Valuable Insights And Significantly Influences Sentiment Classification. Addressing These Gaps, This Thesis Explores Deep Learning-Based Methods For Multimodal Sentiment Analysis, Emphasizing The Semantic Association Between Multimodal Data, Information Interaction, And Color Sentiment Modelling From The Perspectives Of The Multimodal Representation Layer, The Multimodal Interaction Layer, And The Color Information Integration Layer. To Mitigate The Overlooked Semantic Interrelations Between Modalities, The Thesis Introduces "Joint Representation Learning For Multimodal Sentiment Analysis" Within The Representation Layer. This Method, Validated By Rigorous Experiments, Showcases A Marked Improvement In Accuracy, Achieving 76.44% On The Mvsa-Single And 72.29% On The Mvsa-Multiple Datasets, Surpassing Existing Methodologies. In The Multimodal Interaction Layer,
Privacy Preservation Model For Data Exposure In Android Smartphone Usage
(2021-11)
Anizah Binti Abu Bakar
Statistics show there are 6378 million of smartphone users. The usage of mobile applications in smartphones exposes users to privacy risks. This is due to existing works lacking a formalized mathematical model that can quantify both user and system applications risk. There is also no multifaceted data collector tool to monitor user data collection and risk posed by each application. Besides, there is no risk level benchmark that alerts users and distinguishes between acceptable and unacceptable risk levels in smartphone usage. In order to tackle the privacy risk issue, a formalized privacy model called PRiMo is proposed using tree structure and calculus knowledge to quantify the risk in each application, risk posed by each application category, and overall privacy risk faced by the smartphone user.
An Efficient Distributed Slotted Multi-Hop Wireless Mac Protocol For Internet Of Things
(2021-11)
Mohammad Ali Sarvghadi
In Wireless Sensor Networks (WSNs), the Media Access Control (MAC) layer is responsible for coordinating Sensor Nodes (SNs) as their access to the media. Real time data flow needs precise scheduling for sensor nodes to access the medium in order to send the data on time. Time slotted transmission scheduling is needed to access the media without collision. To achieve this, sensor nodes need to be synchronized and know about each other’s notion of time. However, time synchronization requires message exchanges, which incur high overhead. The proposed MAC protocol, which consists of algorithms that minimizes collision and detects hidden terminals, eliminates the message overheads of global time synchronization
Geographical Multicast Disruption Tolerant Networking Mechanism For Internet Of Things
(2022-01)
Wong Khang Siang
Disruption Tolerant Networking (DTN) has been developed to overcome the intermittent connection issue between nodes in areas with poor wireless network connectivity by employing a store-carry-forward paradigm to forward messages to the destination. The existing networking protocols such as Transmission Control Protocol (TCP) and User Datagram Protocol (UDP) are not suitable since there may never be an end-toend path between the sender and the receiver. As the Internet of Things (IoT) devices proliferate, enabling DTN support in IoT environments helps bridge the communication gap between networks with intermittent connectivity such as rural areas and postdisaster scenarios. Group communication is an essential service to enable information exchange and sharing within a group and between groups in such networks. Furthermore, some applications require reliable multicast support over resource-constrained DTN networks. However, there is no well-defined standard for efficient and reliable group communication in DTN. The group communication in a post-disaster scenario that covers a large geographical area presents a more challenging environment for the disaster relief personnel to communicate and coordinate search and rescue missions. A group-based data delivery service is needed in DTN networks with multicast support for communication over multiple geographical areas. In resource-constrained IoT networks, the group-based data delivery needs to be enhanced to provide reliable multicast support for use cases, such as reliable configuration updates.
Landmark Image Discovery Using Network Clustering
(2022-03)
Mohammed Al-Zou’Bi, Ala’A Ahmed
Significant amounts of Internet photo collections are stored online and continue to grow rapidly. This wealth and availability of visual information enable the development of several computer vision applications. Therefore, there is a need for efficient techniques for structuring and organizing this large number of images. In particular, landmark images form a large portion of such collections. Mining of landmark images relies on clustering to group large-scale image collections by the object they depict. The grouping process is a very challenging task due to the variations in the object’s appearance, which can be caused by illumination conditions, differences in scale and imaging viewpoint.
Rule-Based Approach For Detecting Advanced Persistent Threat Using Behavioral Features Of Credential Dumping Technique
(2022-04)
Ali Mohamed, Nachaat Abdelatif
The shift from the manual approach of processing data to the digitized method has made organizational data prone to various attacks by cybercriminals. Advanced Persistent Threat (APT) is a recent threat that has ravaged many industries and governments. APT causes enormous damages for data loss, espionage, sabotage, leak, or forceful pay of ransom money to the attackers. Current security measures of addressing APT attack involve detecting the attacks long after it has happened and failed to provide proactive responses. The current security solutions are deployed to detect APT signature and behaviour after APT bypasses the entire protections and accomplishes lateral movement technique, which makes the current solutions ineffective to resolve APT problem.
Slow Fusion Triplanar Convolutional Neural Networks For Liver Tumor Segmentation
(2022-04)
Chung, Sheng Hung
According to the World Health Organization (WHO) report, liver tumor is one of the leading cause of death in all cancerous disease reported worldwide, with fatalities rate of 745,000 patients in 2014, 788,000 patients in 2015 and 782,000 patients in 2018 respectively. Liver tumor diagnosis and surgery planning are commonly performed with Computed Tomography (CT) scan to assist doctors in evaluating liver tumor and planning of the relevant treatment for the patients. However, there are challenges faced in liver tumor segmentation such as (i) similar intensities between liver tumor and liver tissues, (ii) small and indeterminate liver tumor which are difficult to characterize and (iii) liver tumor with irregular shapes and boundaries. Therefore, an accurate liver tumor detection and segmentation is a crucial prerequisite for liver tumor diagnosis, surgery and treatment planning. In this study, we demonstrate the use of multiple views including axial, sagittal and coronal images as the inputs for Convolutional Neural Networks for liver tumor segmentation, named as Triplanar Convolutional Neural Networks. Our designed network model, Triplanar Convolutional Neural Networks utilize different views of liver CT images to extract discriminative features from the Voxel of Interest (VOI) to classify liver tumor from a healthy liver region in the Liver CT dataset obtained from MICCAI 2017 Liver Tumor Segmentation (LiTS) Challenge.
Incorporating Informative Score For Instance Selection In Semi-supervised Sentiment Classification
(2022-05)
Vivian, Lee Lay Shan
Sentiment classification is a useful tool to classify reviews that contain a wealth of information about sentiments and attitudes towards a product or service. Existing studies are heavily relying on sentiment classification methods that require fully annotated input. However, there are limited labelled text available, making the acquirement process of the fully annotated input costly and labour intensive. In recent years, semi-supervised methods have been positively recommended as they require only partially labelled input and performed comparably to the current preferred methods. At the same time, there are some works reported the performance of semi-supervised model degraded after adding unlabelled instances into training. The contrast of the current literature shows that not all unlabelled instances are equally useful; thus identifying the informative unlabelled instances is beneficial in training a semi-supervised model. To achieve this, informative score is proposed and incorporated into semi-supervised sentiment classification. The experiment compared the accuracy and loss of supervised method, semi-supervised method without informative score and semi-supervised method with informative score. With the help of informative score to identify informative unlabelled instances, semi-supervised models can perform better compared to semi-supervised models that do not incorporate informative score into its training. Although performance of semi-supervised models incorporated with informative score are not able to surpass the supervised models, the results are still found promising as the differences in performance are subtle and the number of labelled instances used are greatly reduced.
A Multi-objective Evolutionary Algorithm Based On Decomposition For Continuous Optimization Using A Step-function Technique
(2022-06)
Chuah, How Siang
Multi-objective optimization is an area of study which solves complex real-world problem that involves two or three objectives. Multi-objective Evolutionary Algorithm based on Decomposition (MOEA/D) is one of the algorithms that utilize the concepts of decomposition and neighbourhood to solve multi-objective problems. One of the recent MOEA/D algorithms, i.e., Constant-distance based Neighbours for MOEA/D with Dynamic Weight Vector Adjustment (MOEA/D-AWACD), integrates the concept of a constant-distance neighbourhood and a dynamic weight vector design. This combination creates a flexible neighbourhood that can adapt to the weight vectors changes. However, MOEA/D-AWACD’s performance is dependent on a constant-distance parameter,
Predictors Of Flourishing Among Elderly In Penang, Malaysia: The Role Of Demographic Variables And Perma Elements
(2022-07)
Liew, Wei Peng
Flourishing is a state of optimal mental well-being. It is the experience of doing and living well in all aspects of one's life, including psychological and social well-being. Flourishing plays a significant role in all life span, and every person has the potential to flourish. Elderly as the growing age group in Malaysia can benefit from being flourished. Nevertheless, current approach to studying flourishing is too general and limited studies have focused on flourishing, particularly among Malaysian elderly from a heterogeneous society with diverse cultural backgrounds. Each ethnicity has its own definition of what it means to flourish. By understanding the definitions of each ethnicity, interventions, programs, or measurements of those who flourish and those who do not can be taken appropriately. As a result, this study focuses on the role of demographic variables (e.g., age, gender, ethnicity, religion, marital status, source of income and education attainment) and elements in PERMA (e.g., engagement, positive relations with others, meaning in life and accomplishment) in contributing to flourishing among Malaysian elderly from various races.
Text Augmentation For Emotion Classification In Microblog Text Using Similarity Scoring Based On Neural Embedding Models
(2022-08)
Yong Kuan Shyang
Emotion classification can benefit from a larger pool of training data but manually expanding the emotion corpus is labour-intensive and time-consuming. Distant supervision can be used to collect large amount of training data in a short period of time using emotion word hashtags, but the collected data may contain excessive noise. In this research, we proposed a text augmentation strategy to efficiently expand the size of positive examples for six emotion categories (happiness, anger, excitement, desperation, boredom and indifference) in EmoTweet-28 by exploiting tweets collected from distant supervision (DS) that are similar to the seed examples in EmoTweet-28 (ET-seed). Similarity scoring approach was used to compute to cosine similarity scores between each DS tweet and all ET-seed tweets under the same emotion category. Seven vector representations (USE, InferSent GloVe, InferSent fastText, Word2Vec, fastText, GloVe, and Bag-of-Words) were experimented to represent the tweets in the similarity scoring approach. DS tweets with high similarity scores were selected to become the augmented instances and annotated with emotion labels. The selection of DS tweets was divided into two categories which are threshold-based selection and fixed increment selection. In addition, we also modified the proposed text augmentation strategy by altering the seed sets used for similarity scoring using clustering and misclassified strategies. All augmented sets were evaluated by training a deep neural network classifier separately to distinguish between the presence or absence of specific emotion in tweets from the test set.
Taylor-Bird Swarm Optimization-Based Deep Belief Network For Medical Data Classification
(2022-08)
Alhassan Afnan Mohammed
Heart disease classification is considered a challenging and complex task in the field of medical informatics. Various medical data classification methods are developed in the existing research works, but achieving higher classification accuracy is a great challenge in the medical sector due to the presence of noisy, and high-dimensional data. Fuzzy clustering-based filtering methods are introduced for essential feature selection. From the selected features, deep learning has become an important stage for disease diagnosis. However, finding the most appropriate deep learning algorithm for a medical classification problem along with its optimal parameters becomes a difficult task. Deep Belief Network (DBN) is a sophisticated learning system that requires a high level of approach and executes well. The major contribution of this research is to introduce a Taylor-Bird Swarm optimization-based Deep Belief Network (Taylor-BSA-based DBN) for medical data classification. Firstly, the pre-processing of medical data is done using log-transformation that converts the data to its uniform value range. Then, the feature selection process is performed using sparse fuzzy-c-means (FCM) for selecting significant features to classify medical data. Incorporating sparse FCM for the feature selection process provides more benefits for interpreting the models, as this sparse technique provides important features for detection and can be utilized for handling high-dimensional data. Then, the selected features are given as input to the DBN classifier which is trained using the Taylor-based bird swarm algorithm (Taylor-BSA). Taylor-BSA is designed by combining the Taylor series and bird swarm algorithm (BSA).

Browse

Browsing Pusat Pengajian Sains Komputer - Tesis by Issue Date

Results Per Page

Sort Options