M.Phil / MS
Permanent URI for this collectionhttps://repository.cuilahore.edu.pk/handle/123456789/36
This collection archives the complete set of theses produced by students of the COMSATS University Islamabad, Lahore Campus.
Browse
8 results
Search Results
Item Machine Learning Approach Towards Motion Planning for Manipulation(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2022) Imran Zahoor Sajid; SP18-RCS-007; LHR TP 8353; Dr. Wajahat M QaziMotion planning is an essential part of robotics. This is important for robots to perform navigational and manipulative tasks. This aims at finding a path from a start position to a goal position while avoiding collisions with obstacles. These days robots are being deployed in public places where they need to deal with uncertainty and a dynamic environment. A motion planning problem generally requires a robot to deal with a dynamic environment, uncertainty, and kinodynamic (velocity, acceleration, and force/torque) constraints. A significant amount of work has been done to solve the problem of motion planning in a dynamic environment. Numerous ways to deal with movement arranging and obstacle avoidance algorithms have been proposed. Those methods can be categorized into sampling-based, imitation learning, and biological inspired, and deep learning. Those approaches have been improved techniques and algorithms for solving motion planning problems. Even though several new efficient techniques have been proposed and many existing ones have been improved, the multitude of motion planning issues has been steadily growing. These problems involve the determination of collision-free paths, modeling of changing environment, real-time recognition of obstacles, and dynamic constraints, etc. These limitations create movement arranging issues really testing and require solid and effective calculations, procedures, and approaches. This study attempts to introduce a deep learning-based planner to Kautham which is a motion planning simulation tool for study and research purposes. Kautham uses OMPL which offers sampling-based planners. Kautham relies on these planners, so it also inherits the problems of sampling-based algorithms. We hope that a deep learning-based planner can reduce the computation time for various environmental settings and can improve Kautham performance and provide a chance to researchers and students to learn and understand deep learning planners. This work results shows that machine learning based planner computation time is less than the sampling based planner.Item Pragmatic Evidence on Android Malware Analysis Techniques: A Systematic Literature Review(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2021) Mr Mian Muhammad Bilal; SP18-RCS-021; LHR TP 7584; Sajid Ibrahim HashmiA large number of state-of-the-art studies on android malware detection and analysis techniques have been published during the last decade. A few comprehensive surveys also exist on the subject. The authors proposed different techniques, tools and frameworks to identify the malware. However, no study attempted to address a systematic review of literature on detection and analysis of android malware methods, systems, and frameworks. In this thesis, we have attempted to present a systematic review of literature on android malware detection and analysis techniques and tools. In this review study, we have identified different android malware detection and analysis methods and tools presented from 2010 to 2021 by following the guidelines of Kitchenham Systematic Literature Review methodology. This thesis presents the 75 most relevant studies out of 3343 published studies. We have determined the android malware datasets used by the android malware detection and analysis techniques. This thesis work has identified that the most used malicious datasets are Genome (39%) and Drebin (36%). We have identified the source code analysis methods in terms of static, dynamic, and hybrid used by android malware detection techniques. We have also identified the limitations and future directions of existing techniques as research gaps for the community. Based on the pragmatic evidence of this research, we have proposed a hybrid analysis based multiple feature analysis framework. This framework will not only address the limitations of static and dynamic based approaches, but it will also analyse evolving android malware datasets using deep neural network and machine learning techniques and improve the accuracy for evolving malware samplesItem Task-Oriented Grasping for Manipulation(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2021) Muhammad Usman Sarwar; SP18-RCS-027; LHR TP 7285; DR. Wajahat M QaziRobot grasping is an important component to perform manipulation tasks. It is considered one of the ways for robots to interact with the environment [1]. Grasp planner plans a valid grasp for the underlying object. A valid grasp is a grasp, which prevents an object to fall [2]. The problem arises when the robot needs to grasp an object for a specific task. In this case, a valid grasp may not be a valid grasp. For instance, a robot needs to pour a liquid from one container to the other. Technically, in this case, a top grasp is a valid grasp but consider the pouring action, the top grasp is not valid. Existing methods to generate a grasp are mostly based on analytical and geometrical solutions. These approaches have performed well in generating a valid grasp, which is not a task/action specific. To generate an action-specific grasp, a robot needs to understand the task and its prerequisites. These prerequisites can be grounded on social norms or technical aspects. These norms and values act as constraints, which robots need to consider while planning a task-specific valid grasp. Induction of task-specific constraints in grasp planning requires the robot to have cognitive skillsets along with geometrical ones. The cognitive skillset requirement includes but is not limited to perception, short- & long-term memories, contextual awareness, situation awareness, semantics, and cultural understanding along with social norms. A cognitive architecture is required to put all these features into an executable framework. The challenge is to design and implement such cognitive architecture in such a way that the robot may be able to learn and perform task-specific grasp considering the given constraints. The research intends to develop such an artifact that allows the robot to perform task-specific grasp. Indeed, at this stage of research, the intention is not to develop a domain-independent task-specific grasp planner. Therefore, the application and validation of the contribution will be validated using kitchen domain objectsItem Detection of Fire and Smoke from Video Sequences(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2020) Ali Hussain; SP18-RCS-003; LHR TP 7280; Dr. Usama Ijaz BajwaSince the introduction of deep neural networks in object detection, fire and smoke has been in the focus of many researchers. The recent state of the art Convolution Neural Network (CNN) based architectures provide more than 95% fire-smoke detection accuracy but in controlled environments, e.g., a fire in server rooms or production lines. Fire burns differently in certain environments; wildfires and domestic fire have distinctive characteristics and burning patterns. A fire detection system should be strong enough to better generalize different fire burning patterns. Training such a system needs a massive amount of annotated data describing unique fire patterns. In this research, a deep neural network-based fire and smoke detection system will be proposed. This network's primary focus will be consuming less training data for better generalization to achieve high accuracy with a low false-positive rate. The results have been reported using the ROC, accuracy, false positive ratesItem Broadcast Latency Minimization Framework for Multimedia Cognitive Radio Networks(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2021) Muhammad Ismai; SP18-RCS-028; LHR TP 7296; Dr. Amjad AliThe wireless multimedia transmissions are increasing substantially because of technological development. The Radio Spectrum (RS) usage is fix and utilized statically around the globe. Cognitive Radio (CR) in wireless multimedia networks has grown in popularity in recent years due to the reduction of the scarcity barrier by sensing the radio-frequency spectrum. This mechanism allows the Secondary Users (SUs) to use the unused bandwidth of the spectrum of the Primary Users (PUs) licensed spectrum over Cognitive Radio Networks (CRNs). Therefore, these multimedia traffic require a spectrum for transmissions. Furthermore multimedia applications require strict Quality of Service (QoS) and this is challenging in CRNs where spectrum is accessed vigorously. The dynamic access techniques in CRNs lead to latency in the network, latency is the time for sharing of perceptive information from source node to faraway node in the network. In this thesis we have developed a framework for minimizing broadcasting latency for multimedia applications. Our framework is composed of seven phases and each phase performs its functionality which lead to the minimum latency. Furthermore the best QoS channel is selected based on fitness function where the fitness values of each channel is calculated and the high fitness value common channel is selected for broadcasting in the system. Our system is efficient system which meets the QoS level satisfaction for multimedia transmissions over a challenging environment where spectrum is not utilized efficiently. The existing results for latency mostly are for non QoS as well as for single radio system and single radio single channel system. Simulation results show that our system is more vigorous and appropriate for supporting minimum broadcast latency over MCRNs.Item Structural based Sentiment Mining for Roman Urdu(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2020) Mubashir Ubaid Ullah; SP18-RCS-013; LHR TP 5986; Dr. Muhammad Waqas AnwarWeb-based data is increasing day by day and plays a vital role in developing people’s opinions. Sentiment mining/analysis is the natural language processing task that helps to identify, classify these opinions. Usually research focus is on resource-rich language for sentiment mining. In this thesis, we performed classification of various sentiments using feature selection techniques for a resource-poor language i.e. Roman Urdu. These classification techniques include chi-square, mutual information and select from model which are implemented on the Roman Urdu Dataset of 11k reviews. Well-known machine learning algorithms are applied for experimental analysis that includes Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), Multinomial Naïve Bayes (MNB) and Multi-Layer Perceptron (MLP). These are applied for character-level & word-level features for n-gram variations that are bigram to 7-gram for character-level classification and Uni, Bi, Uni Bi gram, Uni-Bi-Tri gram & Uni-Bi-Tri-Four gram in terms of word-level classification. Results are being evaluated using accuracy, precision, recall & f1-score. The Highest accuracies for both word-level and character-level achieved are 83.93% and 83.72% which improves the baseline score that was 82.46% on feature union whereas F1-score is 90.51% & 90.42% respectively. Some renowned Neural Network techniques are also applied in this thesis which include CNN, LSTM, & Bi-LSTTM. We achieved maximum results by Bi-LSTM which gives 91.8% accuracy and 91.7% F1-scoreItem Urdu to English Based Unsupervised Machine Translation(Library Information Services, COMSATS University Islamabad, Lahore Campus, 2020) Ahmed Raza; SP18-RCS-005; LHR TP 5983; Dr. Muhammad Waqas AnwarThe process of automatically converting the text from one language to another natural language is Machine Translation. Machine Translation is a subfield of computational linguistic. There are two state-of-the-art machine translation techniques i,e Neural Machine Translation (NMT), and Statistical Machine Translation (SMT). In both techniques, a large corpus is required for the training of the translation model. Urdu counts in low resource languages due to the fewer resources available for computational work. To build a good translation system available resources are not enough. Many languages present in the world have a different structure. Like in Urdu and English, Urdu structure is based on Subject Object-Verb (SOV) and the English structure is based Subject-Verb-Object (SVO). In this study, we presented Urdu to English unsupervised translation model and the practical challenges faced during the work. We try to partially remove the need for parallel corpora and proposed a method to train a Machine Translation System in an unsupervised manner. The proposed system is aimed to provide Urdu to English translation through an unsupervised manner. For this propose, we use Artetxe Author developed a toolkit that is based on Unsupervised Neural Machine Translation (UNMT). This approach tested the models of UNMT which include denoising and on-the-fly back-translation. From denoising model obtain the BLEU score 4.14 and 5.11 for two language pairs UR-EN and EN-UR. From back translation obtain the BLEU score of 5.21 and 6.28 which are better than from the previous score. Back-translation results difference from denoising technique gains +1.07 and +1.17 for two language pairs Urdu to English and English to Urdu. We also faced many challenges during work and effects on pre-processing techniques. Our approach shows promising results in translation of Urdu text into English which is mostly neglected due to its complexitiesItem A Context-Aware Localized Federated Learning Approach for Cloudlet Federation(Library Information Services COMSATS University Islamaabad Lahore Campus, 2020) Sana Latif; SP18-RCS-019; Dr. Syed Asad Hussain; LHR TP 6404Cloud computing (CC) provides storage and computation capability to perform resource intensive Machine Learning (ML) tasks for prediction and decision making. However, cloud services can introduce latency and bandwidth limitations while transmitting data to a remote cloud. Cloudlets are deployed as stand-alone devices at the network edge to bring computation in the closer proximity of users to address cloud computing challenges. Moreover, cloudlets have insufficient resources to train these resource-intensive deep learning models. Therefore, deployment of cloudlet federation can resolve latency, storage, computation, and bandwidth limitations by offloading tasks to a cloudlet within the federation. The selection of a deep learning model to reduce communication and computation cost is a challenge, as cloudlets are not context-aware in terms of network load and latency in the cloudlet federation. Moreover, deep learning model accuracy and prediction results can be affected if end-user devices are unreliable and provide incorrect data for training deep learning models at cloudlets. A cloudlet federation based novel solution is discussed in this thesis for Federated Learning (FL) that monitors network load and resources using a broker. The broker is a centralized entity that will reside within the federation and the global model will be stored inside the broker to make localized decision. COVID-19 X-ray images are used to train the model. The data is divided into 70% training and 30% validation data sets. Besides cloudlets broker will have its dataset to check the validation accuracy of the global model. The sampling data and network parameters such as available storage will be used to extract a context-aware local model for each cloudlet from the generic model. The local model will be converged using Root Mean Square Proportional (RMSProp) based on the mean squared error loss function as it provides higher training accuracy, irrespective of training data size. Convolutional Neural Network (CNN) is used in this thesis with 6 layers for classification of data, one average pooling layer, a dense and a dropout layer. The trained model on each cloudlet is sent to the broker for aggregation. Two aggregation methods are proposed in the thesis. The first one uses the layer aggregation method and the second one is based on the best model selection. Vertical Federated Learning (VFL) is used for model training. Experiments with method one gives a model accuracy of 86% and loss is 14%. However, the best model selection method generates a model ix accuracy of 96% and a loss of 15%. Total time taken for model convergence using conventional FL architecture is approximately 30 minutes. However, the designed architecture takes approximately 28 minutes for completion of global epochs. Based on the results obtained it can be concluded that the designed testbed can identify COVID-19 infected person using X-ray image or CT-Scan with 86% accuracy using FedAvg algorithm and with 96% accuracy using best model selection.