Faculty of IT, Computational Mathematics

Permanent URI for this communityhttp://192.248.9.226/handle/123/15822

Browse

Now showing 1 - 20 of 83

item: Thesis-Abstract
2D Human animation synthesis from videos using generative adversarial neural networks
(2022) Udawatta PN; Fernando S
Synthesizing 2D human animation has many industrial applications yet is currently done manually by animators utilizing time and resources. Therefore, many types of research have been conducted to synthesize human animation using artificial intelligence techniques. However, these approaches lack the quality as well as capability to generalize to various visual styles. Thus, synthesizing high-quality human animations across different visual styles remains a research challenge We hypothesize that given video references for motion and appearance, synthesizing high-quality human animations across a variety of visual styles can be achieved via generative adversarial networks. Here we have come up with the solution known as HumAS-GAN, an acronym for Human Animation Synthesis Generative Adversarial Networks. HumAS-GAN accepts video references for motion and appearance and synthesis 2d Human animations. HumAS-GAN has three main modules, motion extraction, motion synthesis, and appearance synthesis. In motion extraction, the motion information is extracted via pre-trained human pose extraction [21], The motion synthesis module syntheses a motion representation matching the target human’s body structure which is then combined with the human pose coordinates to be used by the appearance synthesis module to generate the Human animation. HumAS-GAN is focused on improving the quality of the animation as well as the ability to use cross-domain/visual-style references to generate animation. This solution will be beneficial for many multimedia-based industries as it is capable of generating high human animations and quickly switching to any visual style they prefer. HumAS-GAN is evaluated against other methods using a custom dataset and a set of 3 experiments designed to evaluate the capability of generating human animations across various visual styles. Evaluations results prove the superiority of HumAS-GAN over other methods in synthesizing high-quality 2d human animations across a variety of visual styles.
item:
A Hybrid approach to natural language machine translation for Sinhala & English
Gray, HP; Karunananda, AS
Machine Translation is one of the least achieved areas in the area of natural language processing. This is because natural languages are complex, a word can have several meanings, a sentence can have several translations and the translation of a sentence may depend on the context. In this report we describe an approach to machine translation for Sinhala and English languages. We postulate that humans are able to translate natural languages through simple rules and experience collected without being knowledgeable about sophisticated language construction such as morphology, syntax, semantics and pragmatic structures. This hypothesis has been inspired by the fact that humans construct word forms, phrases and sentences with new words they learn by using simple rules without even being fully conscious about the rules. We do not ignore the fact that all words in a vocabulary do not follow the same rules for forming words. Humans use specific knowledge about certain words when they construct sentences. Also the word selection in a translated sentence varies depending on the context or the semantics of the sentence. Due to this complexity, we focus on a hybrid approach which uses both rules and statistics. The system described in this thesis focuses on modeling the steps taken by a human to translate a sentence from one language to the other. A bilingual dictionary is used to modal the knowledge of words and synonyms in both languages. Exceptional word dictionaries are used as equivalents to the knowledge of the special words which do not follow the common rules of morphology. The language parsers handle the syntax of sentences in either language. Morphology analyzers are used to handle the rules used in constructing word forms while statistical analyzers are used to handle the proper word usage depending on the syntax. The system was evaluated by comparing human translation with the machine translation output. The two dominating factors considered were, how understandable the translated sentence is and how much information the translated sentence retains compared to the original. The results are up to the expected quality and further work is required to improve the semantics of translation.
item: Thesis-Full-text
An Active self-learning model for deceptive phishing detection
(2023) Ariyadasa SN; Fernando KSD
Phishing presents an ongoing and dynamic threat to Internet users, targeting personal and confidential information. Existing anti-phishing solutions encounter challenges in keeping up with the ever-changing nature of these attacks, leading to performance degradation over time. This study aims to develop an autonomous anti-phishing solution that effectively counters evolving phishing threats through continuous knowledge updates. To address the challenge of detecting the latest phishing attacks, SmartiPhish, an autonomous anti-phishing solution with continuous learning support, is proposed. Utilizing a quantitative research approach, data is collected from trusted third parties at multiple time points to create a valid dataset. The primary outcome is a reinforcement learning solution that leverages a novel deep learning model alongside Alexa rank and community decisions. The innovative use of Graph Neural Networks in the anti-phishing domain, combined with Long-term Recurrent Convolutional Networks, enables SmartiPhish to estimate a website’s phishing probability using URL and HTML content features. Additionally, the study addresses a crucial research gap by developing a reliable method named PhishRepo for collecting and precisely labelling the latest phishing data. SmartiPhish exhibits positive results, achieving a detection accuracy of 96.40%, an f1-score of 96.42%, and an exceptionally low False Negative Rate (FNR) of 0.029. In real-world web environments, the solution outperforms similar solutions and demonstrates enhanced effectiveness against zero-day phishing attacks. Notably, the integration of continuous learning support facilitates a significant 6% improvement in detection accuracy after six weeks. SmartiPhish’s adaptive approach integrates a systematic knowledge acquisition process, enabling dynamic updates of phishing detection features to counter the ever-evolving landscape of phishing attacks. The findings highlight its potential in strengthening cybersecurity measures and provide practical insights for dealing with phishing threats in today’s digital world. Continuously updating its knowledge base, SmartiPhish stands as a strong defence, promising improved protection for Internet users.
item: Thesis-Abstract
Agent based dynamic geographical map navigation system using A* algorithm
(2015-02-22) Weerasinghe, VWARK; Karunanada, Prof. AS
Navigation is becoming rapidly popular in the society with technological enhancement of the world. In many mobile phones, tab pcs, vehicles and so many devices include navigation systems as a default feature. Further, gaming industry and mobile robot industry vastly use navigation technology for various kinds of operations. Due to this popularity, GIS based Navigation systems have become highly influential for day to day life of the society. Though navigation systems are much popular, many GIS based navigation systems have several issues. One of the major issues is, once it calculates and plans path it monitors the selected path only. Because of this reason some paths have become more efficient than calculated path due to traffic situation changes while navigating on the selected route. User misses such kind of paths due to the above issue. Thus the research focus on how to solve above mentioned problem using A* searching algorithm and Agent technology. Fist it discuss in the literature review chapter, different researches conducted related to GIS and robot navigation will be discussed. In the design chapter, gives high level picture about the design. Then gradually it will explain how agents solve this issue by communicating with each other, under implementation chapter. Evaluation chapter discusses how implemented solution has been tested and the other advantages which can be gained by overcoming limitations of the existing release management tools. Under conclusion chapter it discusses whether each objective has achieved by providing appropriate test samples. Achievements of the objectives and problems of the followed process will be discuss in the conclusion chapter by providing appropriate test examples.
item: Thesis-Abstract
Agent Based Dynamic Partitional Clustering
(2015-05-06) Dehideniya, DMMB; Karunananda, Prof. AS
Most of the well established clustering algorithms assume that the underlying clustering structure of data set does not change over time. Hence, those algorithms fail to identify underlying cluster structures in currently available large scale dynamic data sources in an efficient manner. According to the literature, there were many attempts to address this issue by extending well established static clustering algorithms for dynamic context. For instance, Incremental K-Means and Incremental DBSCAN algorithms are two such attempts. Those methods update the model periodically according to changes in the database or change the model parameters whenever new data appear in an incremental manner. Additionally, Fuzzy Logic and Rough Set concepts are also employed to deal with the uncertainty in dynamic data sources. Other than these, Multi-agent technology has been used to address issues in data clustering in both static and dynamic contexts. PADMA and PAPYRUS are two reported Multi-agent clustering systems in 1990s. Moreover, Chaimontree, Atkinson, and Coenen developed a Multi-agent based clustering system which improves the initial cluster configuration by the interaction and negotiation between cluster agents that represent each cluster in the data set. Although there were many attempts to develop agent based clustering algorithms, but there are lesser number of reported works on identification of partitional clusters in a dynamic data source. The study presented in this thesis proposes a Multi-agent based approach to identify partitional clusters in a dynamic data source. Set of partitional clusters in a dynamic data source is identified by interactions and negotiations among the agents who represent data records in the data source. Then identified potential clusters are assigned to what are called Cluster agents. By interactions and negotiations between cluster agents and Data Record agents, the identified cluster configuration is continuously improved according to the internal cluster evaluation measures. The proposed method is evaluated by synthetic data sets with different number of clusters in 2D and 3D spaces. Results indicate that the proposed method successfully identifies the clusters in those data sets with minimal human intervention.
item: Thesis-Abstract
Agent based solution for artificial neural network optimisation
(2015-02-22) Dharmakeerthi, PSM; Karunanada, Prof. AS
Artificial neural networks are highly used in the areas of pattern recognition, feature extraction, function approximation, scientific classification, control systems, noise reduction and prediction. Feed-forward and back-propagation neural networks are the most commonly used artificial neural networks. Many researchers face difficulties when selecting a proper ANN architecture and training parameters. The manual ANN training process is not the best practical solution because it is a much time consuming task. Also most of the people conduct the manual process in an ad-hoc manner without having a proper knowledge about artificial neural networks. At the end of this research project a multi-agent system: MASAnnt (Multi Agent System for Artificial Neural Network Training) was developed to automate the neural network training for feed-forward and back-propagation neural network. Interaction among agents enables emergence of quality training sessions which cannot be archived by an ad-hoc training sessions conducted by humans. It is straight forward to recognize training parameters such as number of hidden layers, number of neurons in each hidden layer, momentum, learning rate, Emax (Error goal) and activate function of an ANN as a set of agents. Inherent features of agents including coordination, communication and negotiation are able to mimic the ANN optimizing and training process by manipulating theses parameters. Our experiments show that the more rational results can be obtained from the system with both simple data sets like XOR as well as with real life data sets. We can conclude that the neural network optimization and training tasks are successfully accomplished by the agent based approach by analysing the results of the evaluation.
item: Thesis-Abstract
Agent based solution to identify the predominant factor for mental disturbance
Chinthaka, VGM; Karunananda, AS
Literature shows that cultivation of cognitive capacities are negatively affected by five major mental factors, namely, Sensory desire, Anger or 111 will, Sloth torpor, Restlessness and Doubt. In many instances they do not appear in isolation, yet as a combination of one or more such factors. Sometimes a factor or more can cause to arise another. This complex behavior results in not being able to exactly determine which one of the factor is dominant. Identifying the dominant mental factor for the disturbance of a person had been a hard task to accomplish since it needs a proper mechanism and a criteria. Yet, it’s essential to treat and overcome the disturbance. Identifying the dominant mental factor for the disturbance is a vital lead and kind of a initiative to few other research areas as well. Therefore research into identification ofthe mental factor that predominantly disturbs a person in his/her studies, daily life and career has become a paramount research interest. A research has been carried out to identify the predominant mental factor for disturbance of an individual by capturing and analyzing Electroencephalography (EEG) brain waves. The research has been conducted to capture EEG wave signals and to train an Artificial Neural Networks for sessions where we exactly know the dominant mental factor. The trained ANN has integrated with a Multi Agent Systems which receives output from ANN for a given EEG waves from of a person in a particular session as percentage values of above mentioned major mental factors, and deliberate on the output generated by the ANN to decide on the most probable. ANN has fourteen inputs which aligns with the sensors of Emotiv EPOC EEG headset and has five outputs which gives percentage values of each mental hindrance that was available in the fed brain wave. Multi agent system consist of five agents representing each mental factor. MAS enhances the result given by ANN and finally come up with the most dominant mental factor for the disturbance of the given brain wave based on mental hindrances. Accuracy of the final result thoroughly depend on data sets which has been used to train ANN and ontology of the agents.
item: Thesis-Full-text
Agent-based solution for improving abstracts
Adhikari, AMTB; Karunananda, AS
Writing abstracts in a comprehensive and meaningful manner is a challenge for any researcher. However an abstract includes limited set of verbs and standard phrases and other good practices of structuring the contents. A research has been conducted to develop an Agent-based Solution for Improving Abstracts. This solution is based on multi agent systems technology and natural language processing together with commonly used verb phrases and other good practices. The system has been developed with nine agents, namely, coordination agent, parser agent, problem agent, solution agent, conclusion agent, content agent, synonym agent, improvement agent and restructure agent. The coordination agent coordinates entire process. The parser agent identifies syntactic information of each sentence and prepares the contents of the abstract for further analysis. The problem agent ensures whether the research problem has been stated in the early part of the abstract and it‘s proportion within the abstract. The solution agent checks for the contents in terms of concepts such as hypothesis, methodology, approach, design, implementation, methods, theoretical framework, technology, hardware, software, and sampling based on the key words. The conclusion agent searches for concepts such as testing, evaluation, data analysis and statistical significance based on the key words. The content agent, improvement agent, synonym agent, and restructure agent are responsible to offer guidelines to modify and improving of the abstract. More importantly, these agents interact with each other and deliberate to reach consensus regarding a solution. For instance, problem agent and solution agent may agree on the proportion of respective contents within the abstract. Each agent has its own Ontology for deliberating with other agents. The Stanford CoreNLP Natural Language Processing Toolkit has been used to develop parser and JADE has been used for development of the entire multi agent system. The system has been developed with JAVA to run on Windows. It has been incrementally tested, and shown interesting results related to checking for completeness of the abstract in terms required materials and suggestion for improvements.
item:
An evolutionary approach to locate urban public services
Punchihewa, TP; Karunananda,AS
An Evolutionary Approach to Locate Urban Public Services postulates how the concept of negotiation in multi agent technology can be used to locate urban public services during city planning. The solution fundamentally comprises of three major categories of public service agents, namely, request, resource and message agents. Once the system is loaded by the human user, terrain data is fed into the system. The terrain agent will be created and draws the city map in the panel. Once the user creates a public service in the city environment, public service agents will be initialized on behalf of them. These, public service agents locate its position in the city, based on the tolerable influence and the inference between them. The system comprises of five modules, geography module, building services module, water services module, natural services module and transportation services module. Geography module handles the terrain related operations in the city environment. Building services module maintains the agent operations of buildings in the city. Water services module handles the operations related to water resources in the city while, natural services module represents agent operations of natural resources. Transportation services module maintains operations related to roads and other transportation resources. Each of the module acts as agents in the multi agent system. All the modules were implemented using Java platform and the agent functionalities were implemented on top of the madkit agent framework. Implemented system was tested to locate different public services under different city conditions. The system was evaluated by providing an evaluator panel an opportunity to build a specific city environment with some public services and to observe the interactions between those public services in the city. Thereafter, their comments about the functionality of the system were obtained and used to enhance the system. The test results reflect that the definition, planning, implementation, testing and documentation of the system had been carried out in an affective and efficient manner. Key Words: Urban Public Services, multi agent systems, madkit agent framework, Java Platform
item: Thesis-Abstract
Applicability of agent technology for software release management
(4/6/2011) Bogoda, BADAS; Bogoda, BADAS
Information Technology industry is one of the most widely spreading industry around the world in recent past as its applicability and adaptability nature for various streams. Influence created by IT industry, on various streams help to accelerate their development in large portion with in a short period of time. Because of quick reaction in problem solving and easy way of storing and retrieving information people tend to replace existing manual systems by computerized systems.// Enterprise level applications are developed as a combination of several components. Modification done in a one component is used to fulfill some functionality of other component. When preparing those modification to deliver for the customer it is required to know which component‘s modification is highly depend on each other and which is not depend on other. Currently this handle by human where developers go through the modification and identify which component is highly depending on the modification done on other components. Developers need to have good communication among each other to identify correct order, the components should arrange. Once the order is identified they start compiling and building jar files. This process is highly time consuming task as there are frequent updates done for the code when fixing issues. This subject to reduce developer‘s effective time he can work. It‘s a big burden for the company as well.// In this thesis it will discuss how to solve above problem by automating the software release management process using Multi-agent technology. In the literature review chapter it will discuss about different researches conducted related to software release management domain. In the design chapter it will give high level picture about the design. Then gradually it will explain how agents solve this by communicating with each other, under implementation chapter. At the evaluation chapter it discuss how implemented solution has been tested and what are the other advantages it can gain by overcoming limitations of the existing release management tools. Under conclusion chapter it discusses whether each objective has achieved by providing appropriate test samples.
item: Thesis-Full-text
Attention monitoring with electroencephalography and artificial neural network
Senarathne, UAC; Karunananda, AS
It’s a well-known fact that people lose attention without notice in many instances. Learning is one of them. If we remain attentive in whole leaning process, it will certainly improve our learning efficacy. If there is any possibility to identify whether we remain attentive during learning process and remind us when we lose the attention, then we can certainly improve our learning effect. In this research, monitoring EEG signals with ANN technology is used to identify whether student remain attentive during learning process. In normal classroom environment, observation is the main way to identify whether student is attentive to the lecture. However, this needs huge effort from teacher to monitor the students. Distance learning is popular among current society, in that case it is rather difficult to use standard methods like observation to monitor the attention. Neurons in our brain are always active and emit electric pulses all the time, hence we can use those to measure the level of attention in above scenarios. A research has been conducted to monitor attention in a particular task by a person and to signal the person immediately so that he/she can get the mind back to the task. The solution will collect the EEG data from subjects and transformed them in to frequency domain using Fast Fourier Analysis (FFT). These data are used to train an Artificial Neural Network (ANN) regarding known EEG wave patterns of attention and monitor the current EEG wave forms in a prescribed time interval. Upon receiving the current wave pattern, it will be fed in to the trained neural network and detect whether the person has lost the attention. Then it will generate a vibration alert to the mobile phone if the attention has been lost. The solution has been tested with in a classroom scenario with 20 students and results shows that 75% of students were able to get back to the class in few seconds.
item: Thesis-Abstract
Automated tourism knowledge graph and intent generation from audio content extracted from videos, by utilizing NLP
(2022) Seneviratne SS; Sumathipala S
Generating a knowledge graph for a chatbot is a time-consuming exercise which needs the help of an expert relevant to the field. This thesis presents our approach to synthesizes the creation of a knowledge graph and intents for a chatbot. Currently, the creation of a knowledge graph and intents for a chatbot is a tedious process and this process does not extract data from videos. Developing a chatbot also requires the support of experienced software engineers. This platform allows a user to build a customized chatbot according to a specific requirement in any field, without the intervention of experts. It also allows for the seamless development of a comprehensive knowledge graph from the video content through a simple and less tedious approach. The platform uses Natural Language Processing (NLP) machine learning models such as Naive Bayes and Logistic Regression and grammar correction techniques to supplement the experience of the users. The working process of this proposed system is Knowledge Extraction and generating the Knowledge Base. The user inserts keywords related to the chatbot’s domain as the first step of the process. The system retrieves the search results from YouTube. Finally, NLP will be used to retrieve data contained in videos to create a preliminary knowledge graph and intents for a chatbot. A scheduler is then activated automatically from time to time to update the knowledge graph and intents. The knowledge graph and intents generated have been tested on a chatbot created using the Rasa framework, with the chatbot giving the correct answers when questioned by a user.
item: Thesis-Abstract
Automating cephalometric analysis in orthodontics using artificial neural network
Ariyarathna, GDWM; Karunananda, AS
This study presented an Artificial Neural Network approach to promote Automate Cephelamatric Analysis in Orthodontics. Analysis and interpretation of standardized radiographs of the facial bones have become an important clinical task in Orthodontics. Conventional method of locating Landmarks depends on manual tracing ofthe radiographs. Since this is time consuming and error proven, demand for completely automate analysis and diagnostic tasks have increased. This study has critically reviewed four major problems in Cephelamatric Analysis; Precision of Landmark identification, Enormous time consumption, Subject to human errors and Need of continues support from experts. We argue that, issue of lack of autonomous solutions for Cephelamatric Analysis has been claimed to be the main problem with conventional approaches. There have been previous endeavors to Automate Cephalometric Analysis using Hand Crafted Algorithms, Mathematical or Statistical Models and Artificial Intelligence techniques. In any case accuracy was the same or worse than the one of manual identification. Therefore the aim of this investigation was to propose an Artificial Neural Network approach to computerize the Cephalometric Analysis. It is evident from the literature that, Neural Networks can introduce very high level of autonomy and accuracy in modeling real world problems. Therefore we hypothesized; Cephalometric Analysis can be automating by using self organizing feature of ANN. The proposed system automates Cephalometric Analysis along four dimensions. I.e. Image Acquisition using a Cephelostast and a scanner in order to capture the images and scan the images. Image Processing and Computer Vision to perform diffusion on gray scaled images and to detect possible edges using Canny. Two Landmarks, point-Me by finding the first existent edge ofthe image from RHs to LHS and edge starting from ‘Me’ is ended suddenly from the point -UIT , have identified and localized during this module. Coordinate along to the downward values of remaining extracted edges used as input to the ANN to detect other landmarks which cannot be identified directly during Computer Vision. Classify landmarks according to their geometrical specifications using a Competitive Neural Pinpoint the land marks according to the mean value of each cluster Network. obtained during ANN training. Users of the system are Orthodontists who will be benefitted from high level of accuracy and relatively fast outputs.
item: Thesis-Abstract
Automating e-Channelling in Mobile Platform using Multi-Agent Technology
(2015-10-20) Chandrapala, HM; Karunananda, A
All human beings wish to lead a healthy life irrespectively of their race, religion, age or social status. Accordingly they temp to seek medical advice or treatment for immediate relief when they feel or sense that they suffer from any kind of illness. They need to consult a doctor in this regard. There are many ways to consult a doctor. The e-channelling has become the most popular way among the patients. Nevertheless, those approaches are time consuming and costly. In these approaches, the process iscomplicated, as the patient or the user has to find out more information from various sources and synthesize them manually tocomplete the e-Channelling process. The main objective of this research is to overcome those identified issues by using the multi agent technology and the mobile technology. i.e., giving more autonomy and reducing the human intervention using the multi agent approach. Sincethe intelligent agents have the feature of autonomy, the human intervention will be reduced and as a result system will become more efficient. Presently, the mobile phone is a very common device andthe majority of the population use the mobile phones. Therefore, if we can introduce the mobile phonesalso to this system, the process will be more effective and with the few taps on the mobile phones, the answer will be displayed. The proposed system has nine agents. They are, the - Message Space Agent, the Hospital Location Agent, the Disease Agent, the Data Agent, the Appointment Managing Agent, the Hospital Checking Agent, the Best Doctor Selection Agent, the Mobile Interfacing Agent and finally the agent that resides on the user’s mobile device. The usercan use the doctor’s name, the hospital, the category and the disease as inputs. However all the above inputs are not mandatory. If the user wants to search a doctor via disease, only the disease name is required. This same theory will be applicable for the category as well. The salient feature of this system is, that it is able to determine the recommended doctor for the particular category or the disease. Therefore, the user does not require to have any idea or knowledge about the best doctor or the recommended doctor for the particular category or the disease. The core agents in the proposed system are developed by using the JADE. The mobile part of the system is developed by using the JADE-LEAP. The proposed system is evaluated by thirty participants and the results were gathered. The results show that the proposed system is more automated than the current system. Furthermore, it shows that the time and the cost are reduced. Accordingly, the new system will address most of the issues relating to the current system.
item: Thesis-Abstract
Autonomous solution for design of curriculum
Samarasinghe, SADI; Karunananda, Prof. AS
Curriculum design is a major research challenge in the rapid changing world of knowledge. The curriculum represents the expression of educational ideas in practice and it includes all the planned learning experience of a school or educational institution which is interconnected with each other. Therefore curriculum design is complex process which takes place by negotiation within committee of experts in the field. Due to the interconnectivity of its different aspects it is time consuming task to design curriculum within a complex environment. As per these reasons conventional software solutions were not able to automate the process of curriculum designing to meet dynamically changing requirements in a complex environment. However some modern approaches proposed intelligent solutions for the area of curriculum development. These solutions were limited to few areas in the curriculum and do not provide better solution within the complex environment. Therefore this system provides an autonomous solution for curriculum designing within an interconnected environment with the use of ontology and multi-agent technology. This solution mainly focuses on curriculum developers in higher education. The user can input a curriculum and upon the request curriculum agents such as credit agent, subject agent, prerequisites agent are created and agents obtain knowledge through the curriculum ontology. Agents communicate and negotiate with each other and autonomously produce the well balanced curriculum according to the given input. The system consists of two major modules for user interaction, process formulation. The User Interaction module has been designed to interact with the user and this module is mainly process with the Request Agent. Process Formulation is based on MAS module and it has several resource agents such as Credit Agent, Pre-Requisite Agent, Subject Agent and the message space. When the Request Agent sends requests through the message space, related agents are activated and attend the specific task. Curriculum ontology provides the knowledge needed to create a well balanced curriculum. The entire multi agent system is developed based on the Java Agent Development Environment. Ontology is developed using Protégé ontology editor. The system has been evaluated using curriculums in the Faculty of Information Technology, University of Moratuwa. Keywords- Multi Agent System, Ontology.
item: Thesis-Abstract
A Big data analytic framework over federated data centers for intelligent travel recommenders
(2021) Udayanthi HPI; Silva T
Big Data is a series of enormous and complex data sets that are nearly impossible to store and process using traditional data storing and processing methods. The emergence of heterogeneous data in different domains causes significant challenges in data manipulation and decision making. In recent years, the requirement for analysis of heterogeneous data on distributed data storages has been increased and has gained a lot of researchers’ attention. Distributed data storage systems and parallel data processing techniques are typically used for data-intensive computing today. Due to the rapid growth of data, a single-cluster environment is inadequate to process that much data. At the same time, there are heterogeneous data sources on different platforms, which need to inter-connect to derive meaningful analysis. The MapReduce software paradigm has surfaced to fill the gap, and it has been successfully operating on systems. However, only single cluster environments are supported by the current implementation of MapReduce and cannot be applied to federated heterogeneous data centers. Hence, it does not have enough capabilities to process heterogeneous data sources. This research presents a big data analytic framework that supports the integration of heterogeneous data sources on distributed computing models across different data centers. The architecture of this framework is based on a master/slave distributed computing model and Map - Reduce - Merge - GlobalReduce is presented as the programming model. Besides, the performance of the novel approach is measured under different cluster configurations, and experimental evaluations had shown promising results for the proposed framework compared to a single cluster environment.
item: Thesis-Full-text
Biometric authentication system using multi-agent technology in border control
(2020) Thenuwara SS; Karunananda AS
In today’s world security is the most important aspect. Border criminals, frauds, unauthorized immigrants are burning issues in Sri Lanka within the last few years due to the lack of proper identification system i.e. duplicate passports, fake identity, etc. Therefore, the efficiency and accuracy of the traditional authentication system are not good enough to overcome this disaster. Furthermore, cryptanalysis and brute force attacks are dramatically strong with uncontrolled demanding of computational power. In fact, the efficiency and accuracy of the authentication are not enough to cater the future authentication systems by comparing traditional user authentication techniques. Therefore, Biometrics is the ideal solution for authentication as it has advantages over conventional systems. Second, it's not important to recall a biometric and it can't easily be lost. It makes the client much smoother. In addition, it is not easy to stolen or loan a biometric to a relative. Biometrics can provide greater security and comfort than traditional methods of human identification. Even if we don't want to replace a conventional method with a biometric one, we are certainly future consumers of these systems, which will even be mandatory for new passport models. Therefore, it is important to be familiar with biometric security engineering possibilities. The most common way for people to perform biometric authentication maybe facial recognition. Face recognition can be based on single images, multiple still images or video sequences. Although most of the efforts have been traditionally dedicated to the former, the latest is rising, probably due to the lowering of prices in devices for image and video acquisition. The system was designed using the technology of multi-agents and it has two phases biometric capturing phase at VISA granting process and biometric recognition phase at border points. The product's longevity is maintained through the use of biometric and nonbiometric techniques. This solution has been implemented as a multi-agent program with a hybrid approach that identification of faces and validation of fingerprint. The system has been tested in a border protection environment which is a more time-critical real-world application and notices the 87% accuracy in the recognition phase. The proposed system evaluated with traditional border management software and time calculated for each participant. Those participants have given informed consent participation at the evaluation.
item: Thesis-Full-text
Brian Computer Interfacing for Game Controlling
Thomas, TID; Karunananda, AS
The way human interacts with electronic games has changed in a dramatic manner in the past decade. Many individuals utilize mouse, keyboards, joysticks and even motion sensors in order to provide input commands to various computer gaming applications. Nonetheless, above mentioned modes of interaction has its confines when it comes to individuals with physical limitations such as handicapped and paralyzed. Thus, addressing this problem Neurogaming has become a hot topic in the past few years where brain wave signals namely electroencephalogram signals are used to control different aspects of an electronic gaming application, this approach eliminates the traditional hand coordinated interacting mechanisms allowing even a handicap to interact with a gaming interface with ease. The main objective of the research has been to implement a system which enables usage of an available electroencephalogram device for instance NeuroSky mind wave mobile, in order to aid the individuals with physical limitations and also to provide near real time attention input, incorporating all parts of a functioning brain computer interface system. These parts are 1) acquiring the electroencephalogram signal 2) process and classify the electroencephalogram signal to extract the attention level and 3) use the attention to control a feature in a multi agent game. This thesis report outlines the step-by-step design of the attention racer system which incorporate module level design and interactions among various components of the system. Furthermore, the implementation details of the attention racer system covers the core code segments and the flow of the system. The implemented system was evaluated using 15 participants. Initially they had to undergo through an attention span test and individual who scored more than 75% was selected for the second phase of the evaluation which was training data acquisition and attention racer valuation. After the attention span test 10 individuals were selected for final evaluation where they were evaluated for attention in different environments. The evaluation results asserted that system worked with 70% accuracy in detecting human attention level and using it to control the speed of the car.
item: Thesis-Full-text
Capsule network based super resolution method for medical image enhancement
(2020) Munasingha SC; Fernando S
Medical imaging has been one of the most attentive research and development areas since the 1950s, particularly due to the contribution to disease diagnosis. Despite the fact that imaging technologies have been advanced in multiple ways, yet resolution limitations can be observed. To overcome the resolution limitations, various image enhancement techniques have been used. Image Super-Resolution (SR) is the latest technique in the list to achieve higher resolution with much lower resolution images. Earlier, frequency based and interpolation based SR techniques were used for SR. The afterward achievements in SR techniques are obtained via Convolution Neural Network (SRCNN) based methods and have several flaws. Capsule net (Caps Net) is the state of the art alternative methodology for the problems which were previously solved by CNN. One recent attempt was made to assess the Caps Net for SR task. This new area has a lot to be explored. Especially the time inefficiencies of this approach should be addressed along with accuracy improvements. In this research several capsule network routing mechanisms have been investigated for Super Resolution pipeline with a medical image dataset. Standard Dynamic Routing and Expectation Maximization Routing methods are re-configured to improve the accuracy. Above all, a novel integration of state of the art routing mechanism, Inverted Dot Product based Attention Routing mechanism is introduced for Super Resolution task. With 300,000 medical image training pairs and 2,500 evaluation pairs, every model was evaluated. Along with different image quality indexes, it was shown that the Dynamic Routing based method outperformed all methods and the newest Attention Routing based approach has shown similar image quality performance to that of the state of the art method FSRCNN and less time complexity to that of the existing Caps Net based approaches. This implies that clinicians can use this system effectively in a clinical setting.
item: Thesis-Abstract
Co-word analysis based automatic web search
(7/24/2012) Bambarasinghe, BANM; Karunananda, AS
Automatic searching, knowledge acquisition and question answering are crying needs among the contemporary World Wide Web users. However conventional web is the major barriers for realizing above applications. This is because almost all important information in it is in natural languages and natural languages are very hard to be manipulated and understood by a computer. As a solution, more than a decade ago, semantic web was introduced, there was a lot of hope on machine understandability of the web. However the semantic web is still very far from realization due to the effort required for semantic tagging of available information. In this project we try to build a solution for automatic searching in conventional web and similar information sources by mimicking the human natural language learning and knowledge representation process. Our approach is based on the hypothesis, which inspired by popular philosophy of science, that learning is matching known facts with new facts. In the context of conventional web, we employ statistical natural language processing technique co-word analysis for matching already available facts with new facts collected during automatic searching process. As a proof of above hypothesis we have built a personalized automatic knowledge extraction application. That extracts knowledge from conventional web or similar information source regarding user queries and present synthesized documents related to the knowledge area of the query. Evaluation done by manual comparison of documents produced by the application and a document produced by a human user by web searching. Results showed automatic knowledge acquisition performs acceptable manner in most of the situations.