Computer Sciences Commons^™

Regularized Deep Network Learning For Multi-Label Visual Recognition, Hao Guo

Theses and Dissertations

This dissertation is focused on the task of multi-label visual recognition, a fundamental task of computer vision. It aims to tell the presence of multiple visual classes from the input image, where the visual classes, such as objects, scenes, attributes, etc., are usually defined as image labels. Due to the prosperous deep networks, this task has been widely studied and significantly improved in recent years. However, it remains a challenging task due to appearance complexity of multiple visual contents co-occurring in one image. This research explores to regularize the deep network learning for multi-label visual recognition.

First, an attention concentration …

Deep Learning Based Sound Event Detection And Classification, Alireza Nasiri

Theses and Dissertations

Hearing sense has an important role in our daily lives. During the recent years, there has been many studies to transfer this capability to the computers. In this dissertation, we design and implement deep learning based algorithms to improve the ability of the computers in recognizing the different sound events.

In the first topic, we investigate sound event detection, which identifies the time boundaries of the sound events in addition to the type of the events. For sound event detection, we propose a new method, AudioMask, to benefit from the object-detection techniques in computer vision. In this method, we convert …

Deep Learning Based Models For Classification From Natural Language Processing To Computer Vision, Xianshan Qu

Theses and Dissertations

With the availability of large scale data sets, researchers in many different areas such as natural language processing, computer vision, recommender systems have started making use of deep learning models and have achieved great progress in recent years. In this dissertation, we study three important classification problems based on deep learning models.

First, with the fast growth of e-commerce, more people choose to purchase products online and browse reviews before making decisions. It is essential to build a model to identify helpful reviews automatically. Our work is inspired by the observation that a customer's expectation of a review can be …

Detecting The Intent Of Email Using Embeddings, Deep Learning And Transfer Learning, Zaid Alibadi

Theses and Dissertations

Throughout the years' several strategies and tools were proposed and developed to help the users cope with the problem of email overload, but each of these solutions had its own limitations and, in some cases, contribute to further problems. One major theme that encapsulates many of these solutions is automatically classifying emails into predefined categories (ex: Finance, Sport, Promotion, etc.) then move/tag the incoming email to that particular category. In general, these solutions have two main limitations: 1) they need to adapt to changing user’s behavior. 2) they require handcrafted features engineering which in turn need a lot of time, …

Programimi Imperativ, Reaktiv Dhe Funksional, Aulona Avdyli

Theses and Dissertations

Me zhvillimin e programimit ndër vite, mund të dallojmë shumë e shumë fjalë dhe terma të reja çdo ditë, duke na ngritur dhe ngjallur kureshtjen të lexojmë dhe të dimë më shumë rreth tyre, pasiqë vet profesioni që kemi zgjedhur është i tillë. Ndër to programimi Imperativ, Reaktiv dhe Funksional, tri lloje të programimit, të cilët ndoshta secili nga ne i kemi përdorur, por pa e ditur saktë me cilin nga ta jemi duke punuar, duke i veçuar në bazë të emrit dhe karakteristikave të tyre.

Në bazë të këtij punimi të diplomës, unë do bëj përpjekje të mund të …

Krahasimi I Teknologjive H.264 Dhe H.265 Në Kompresimet Audio-Video Materialeve, Milot Hoti

Theses and Dissertations

Teknologjia e kompresimit të audio dhe video materialeve, për qëllim primar ka që secili përdorues t’i qaset atij materiali që ka nevojë ta shikojë. Pa teknologjinë e kompresimit kjo mundësi do të ishte shumë e kufizuar dhe jo efikase. Përmes kësaj teknologjie përfitojnë përdoruesit dhe kompanitë të cilat ofrojnë shërbimin për shikim të video materialeve. Përfitimi tek përdoruesit nisë, që nga mundësia e ruajtjes së një fotografie të ditëlindjes së parë, e pse jo edhe incizimi i fryrjes së qirinjve në atë ditëlindje. Ndërsa kompanitë shfrytëzojnë këtë teknologji për ruajtjen dhe shpërndarjen e materialeve, tek secili që dëshiron ta shohë …

Personalized Learning Path Generation Based On Genetic Algorithms, Lumbardh Elshani

Theses and Dissertations

A substantial disadvantage of traditional learning is that all students follow the same curriculum sequence, but not all of them have the same background of knowledge, the same preferences, the same learning goals, and the same needs. Traditional teaching resources, such as textbooks, in most cases orient students to follow fixed sequences during the learning process, thus impairing their performance. Curriculum sequencing is an important research issue for learning process because no fixed learning paths will be appropriate for all learners. For this reason, many research papers are focused on the development of mechanisms to offer personalization on learning sequences, …

Aplikimi I Rrjetave Kompjuterike Dhe Cloud Sherbimeve Ne E-Health Care, Ilir Gojani

Theses and Dissertations

Gjithëkush flet për eHealth këto ditë, por pak njerëz kanë dalë me një përkufizim të qartë të këtijë termi relativisht të ri. Pak në përdorim para vitit 1999, ky term tani duket se shërben si një fjalë e përgjithshme që përdoret për të karakterizuar jo vetëm "mjekësinë e digjitalizuar", por gjithashtu pothuajse çdo gjë që lidhet me kompjuterët dhe mjekësinë. Termi eHealth është krijuar dhe përdorur duke u bazuar në fjalët tjera “e-fjal” si për shembull “e-commerce, e-business, e-mail”, me të cilen fjalë do të mund ti referoheshim mundësive të reja që interneti po i hap në fushën e kujdesit …

Përformanca Dhe Efikasiteti I Arkitekturës Shumë Shtresore Në Iot Edge, Donjetë Hyseni

Theses and Dissertations

Internet of Things (IOT) është bërë një teknologji kyqe e cila lidh së bashku pajisje të ndryshme me funskione të ndryshme. Është një teknologji e cila mundëson shërbime dixhitale të disponueshme kudo për përdoruesit e fundit bashkohorë.

Prandaj me ardhjen e Internet of things (IOT) dhe konvergjencen e të dhënave duke përdorur shërbime cloud,computing e të dhënave është shtyrë në horizonte të reja.Një model I ri I computing,edge computing proceson të dhënat në edge të rrjetit shumica e të dhënave të gjeneruara në edge të rrjetit drejton kërkesen për një kohë të lartë të përgjigjies.Në këtë punim u diskuta për …

Dokumentimi I Arkitekturës Softuerike, Genita Abazi

Theses and Dissertations

Dokumentimi ka luajtur një rol kyç në të ndihmuarit e kuptimit të softuerit për një kohë të gjatë. Dokumentimi softuerik mund të shprehet në forma të ndryshme, varësisht nga audienca dhe natyra e softuerit. Mirëpo jo gjithmonë dokumentimi e ilustron operimin e softuerit përkatës në mënyrë të saktë, si rezultat i së cilës softueri ka potencial të dështimit brenda një kohe shumë të shkurtër.

Ky punim diplome adreson çështjet lidhur me rëndësinë e posedimit të dokumentimit të softuerit, jep një prezantim të shkurtër të standardeve të dokumentimit, të cilat ofrojnë një hyrje dhe udhëzime të përgjithshme për dokumentim të mirëfilltë …

Recommender Systems Using Cloud Technology, Gent Jashari

Theses and Dissertations

Ueb aplikacionet që ofrojnë sistemet e rekomanduara tani më janë bërë trend që çdo uebfaqe t’i ketë në mënyrë që ta tërheqin përdoruesin dhe “mos ta lejojnë atë që të kalojë pa e klikuar sugjerimin e rekomanduar”. Kjo është arritur duke përdorur teknika të ndryshme si: Collaborative filtering, Content-Based filtering dhe Hybrid filtering, ku kjo e fundit ndihmon për tejkalimin e sfidave që na shfaqen në sistemet e rekomanduara.

Përdorimin e cloud shërbimeve na ofrojnë shumë kompani mirëpo ndër ato më të njohurat janë: Microsoft Azure, Amazon Web Services dhe Google Cloud. Studimet e shumta tregojnë që për të krijuar …

Find-Doctor Application, Kastriot Beha

Theses and Dissertations

Nëpërmjet kësaj ideje unë synojë të krijojë një web aplikacion i cili u mundëson përdoruesëve të kërkojnë mjekë të departamenteve të ndryshme sipas nevojave të secilit përdorues. Renditja e mjekëve pas kërkimit nga ana e përdoruesve apo pacientëve bëhet përmes rating(vlerësimit) nga doktori më i mire pra doktori me rating me te lartë deri tek doktori me rating më të ulët. Përveq renditjes së mjekëve ne ofrojm edhte statistika apo charts për të treguar top 5 mjekët më të mire, në mënyrë qe përdoruesve të ju lehtësojm gjetjen e doktorit të caktuar dhe të duhurit për nevojat e tyre. Poashtu …

Gjuha Deklarative Dhe Koha E Ekzekutimit (Declarative Language And Runtime Systems), Valmir Mujku

Theses and Dissertations

Ambienti që na rrethon përbëhet nga shumë pajisje me diversitet komponimesh por nuk mund të themi se një shumicë prej tyre nuk i përkasin botës së teknologjisë. Evolucioni i çdo produkti teknologjik është përcjell me një seri te problemeve dhe çështjeve që janë paraqitur gjatë fazave të ndryshme të planifikimit, zhvillimit dhe menaxhimit. Gjatë fazave zhvillimore, saktësisht ndërtimit te sistemeve softuerike apo programeve aplikative është paraqitur nevoja e depozitimit të të dhënave pavarësisht a kanë qenë si source code (kod burimor) apo të dhëna që janë akumuluar gjatë përdorimit të sistemit. Depozitimi i të dhënave bëhet në bazat e të …

Konsiderata Për Menaxhimin E Internetit Të Gjërave (Iot), Siguria Në Internet Dhe Rreziqet E Privatësisë, Zeqir Nika

Theses and Dissertations

Interneti i Gjërave (IoT) është një koleksion i shpejtë evoluues dhe zgjerues i teknologjive të ndryshme që ndërveprojnë me botën fizike. Shumë organizata nuk janë domosdoshmërisht të vetëdijshëm për numrin e madh të pajisjeve IoT që ata janë duke përdorur tashmë dhe se si pajisjet IoT mund të ndikojnë në sigurinë në internet dhe rreziqet e intimitetit ndryshe nga sa bëjnë pajisjet e teknologjisë konvencionale (IT).

Qëllimi i këtij publikimi është të ndihmojë organizatat të kuptojnë dhe menaxhojnë më mirë rreziqet e sigurisë në internet dhe rreziqet e intimitetit të lidhura me pajisjet individuale të Internetit të Gjërave (IoT) gjatë …

Menaxhimi I Detyrave Dhe Orarit, Azdren Bytyqi

Theses and Dissertations

Duke analizuar vështirësitë e procesit të punës në përgjithësi, rendësinë e mbarëvajtjes së saj dhe kompleksitetin e saj, kjo temë synon që të ofrojë një zgjidhje që sado pak ta lehtësojë menaxhimin e këtij procesi. Në çfarë kohe punëtori shfaqet në punë, e kur del nga puna, detyrat e caktuara për të dhe sa i realizon ato punëtori, pajisja e tij me kontratë, këtyre, e problemeve të ngjashme me to, progarmi i krijuar, synon t’u japë zgjidhje. Rëndësia e kësaj teme dhe asaj që kjo temë shtjellon qëndron në atë se shpesh herë kemi parë keq shfrytëzime si nga ana …

Remote Monitoring Of Memory Data Structures For Malware Detection In A Talos Ii Architecture, Robert A. Willburn

Theses and Dissertations

New forms of malware, namely xC;leless malware and rootkits, pose a threat to traditional anti-malware. In particular, Rootkits have the capacity to obscure the present state of memory from the user space of a target machine. If thishappens, anti-malware running in the user space of an axB;ected machine cannot be trusted to operate properly. To combat this threat, this research proposes the remote monitoring of memory from a second, secure processor runningOpenBMC, serving as a baseboard management controller for a POWER9 processor, which is assumed vulnerable to exploitation. The baseboard management controller includes an application called pdbg, used for debugging …

A Framework For Autonomous Cooperative Optimal Assignment And Control Of Satellite Formations, Devin E. Saunders

Theses and Dissertations

A decentralized, cooperative multi-agent optimal control framework is presented to offer a solution to the assignment and control problems associated with performing multi-agent tasks in a proximity operations environment. However, the framework developed may be applied to a variety domains such as air, space, and sea. The solution presented takes advantage of a second price auction assignment algorithm to optimally task each satellite, while model predictive control is implemented to control the agents optimally while adhering to safety and mission constraints. The solution is compared to a pseudospectral collocation method, and a study on tuning parameters is included.

Cyclic Pursuit, Daniel E. Oke

Theses and Dissertations

This thesis analyzes cyclic pursuit with the intent of developing swarm attack strategies for autonomous agents. Research was focused on finding the effects of pursuers capture range, evader speed and size of formation on the probability of escape. The temporal evolution of several polygonal formations was analyzed. The polygons could be regular or arbitrary polygons. The thesis demonstrated that an increased capture range, formation size, reduced evader speed aided capture probability. Irregular n-gon formations reduced to n-1 gon repeatedly, pursuer clusters formed until two clusters remained which eventually came together, so all the n pursuers coalesced until convergence. Regular n-gon …

Error Detection In Quantum Algorithms, Simeon R. Hanks

Theses and Dissertations

Quantum computers need to be able to control highly entangled quantum states in the presence of environmental perturbations that lead to errors in calculations. Progress in superconducting qubits has enabled the development of computers capable of running small quantum circuits. The current era of Noise Intermediate Scale Quantum computing has a high error rate. To alleviate this error rate we apply an encoding scheme that allows us to remove results with known errors improving the quality of our results. The encoding uses multiple qubits as a single logical qubit and balances the natural tendency of state-of-the-art quantum computers to decohere …

Aircraft Inspection By Multirotor Uav Using Coverage Path Planning, Patrick H. Silberberg

Theses and Dissertations

All military and commercial aircraft must undergo frequent visual inspections in order to identify damage that could pose a danger to safety of flight. Currently, these inspections are primarily conducted by maintenance personnel. Inspectors must scrutinize the aircraft’s surface to find and document defects such as dents, hail damage, broken fasteners, etc.; this is a time consuming, tedious, and hazardous process. The goal of this work is to develop a visual inspection system which can be used by an Unmanned Aerial Vehicle (UAV), and to test the feasibility of this system on military aircraft. Using an autonomous system in place …

Display Design To Avoid And Mitigate Limit Cycle Oscillations (Lco) On The F-16c, David J. Feibus

Theses and Dissertations

The U.S. Air Force F-16 Fighting Falcons flying characteristics and flight envelope are dynamic and defined by its external weapon stores configuration. The employment of its munitions at certain speeds can put the F-16 into a flutter-like state in which Limit Cycle Oscillations (LCO) are induced. In LCO, a pilots fine motor control might be hindered, and the aircraft may lose combat effectiveness until flight conditions are reduced. The current research attempted to provide pilots with a predictive feedback display to avoid an LCO-susceptible configuration by increasing their situation awareness about the consequences of employing certain munitions to their flight …

Correction Of Back Trajectories Utilizing Machine Learning, Britta F. Gjermo Morrison

Theses and Dissertations

The goal of this work was to analyze 24-hour back trajectory performance from a global, low-resolution weather model compared to a high-resolution limited area weather model in particular meteorological regimes, or flow patterns using K-means clustering, an unsupervised machine learning technique. The duration of this study was from 2015-2019 for the contiguous United States (CONUS). Three different machine learning algorithms were tested to study the utility of these methods improving the performance of the CFS relative to the performance of the RAP. The aforementioned machine learning techniques are linear regression, Bayesian ridge regression, and random forest regression. These results mean …

Long Distance Bluetooth Low Energy Exploitation On A Wireless Attack Platform, Stephanie L. Long

Theses and Dissertations

In the past decade, embedded technology, known as the Internet of Things, has expanded for many uses. The smart home infrastructure has drastically grown to include networked refrigerators, lighting systems, speakers, watches, and more. This increase in the use of wireless protocols provides a larger attack surface for cyber actors than ever before. Wireless loT traffic is susceptible for sniffing by an attacker. The attack platform skypie is upgraded to incorporate Bluetooth Low Energy (BLE) beacon collection for pattern-of-life data, as well as device characteristic enumeration and potential characteristic modification. This platform allows an attacker to mount the skypie to …

Enumerating And Locating Bluetooth Devices For Casualty Recovery In A First-Responder Environment, Justin M. Durham

Theses and Dissertations

It is difficult for first-responders to quickly locate casualties in an emergency environment such as an explosion or natural disaster. In order to provide another tool to locate individuals, this research attempts to identify and estimate the location of devices that would likely be located on or with a person. A variety of devices, such as phones, smartwatches, and Bluetooth-enabled locks, are tested in multiple environments and at various heights to determine the impact that placement and interference played in locating the devices. The hypothesis is that most Bluetooth devices can be successfully enumerated quickly, but cannot be accurately located …

A Fixed-Point Method For Computing Steady-State 2d Laser-Fluid Interactions, Jeremiah S. Lane

Theses and Dissertations

This research introduces a fixed-point numerical approach for solving the steady-state Navier-Stokes (NS) equations on a finite two-dimensional (2D) domain. The steady-state interaction between a high energy laser beam and its surrounding fluid medium is important to researchers in the field of high energy laser beam propagation. The solutions to the steady-state Navier-Stokes equations provide a model for uncovering the steady-state behavior of the fluid medium, which is useful for the modeling of thermal blooming in laser beam propagation. Numerical solutions remain the only tenable option for solving the NS equations, wherein numerical speed and fidelity beget the utility of …

Formal Verification For High Assurance Software: A Case Study Using The Spark Auto-Active Verification Toolset, Ryan M. Baity

Theses and Dissertations

Software is an increasingly integral and sophisticated part of safety- and mission-critical systems. Poorly written software can lead to information leakage, undetected cyber breaches, and even human injury in cases where the software directly interfaces with components of a physical system. These systems may range from power facilities to remotely piloted aircraft. Software bugs and vulnerabilities can lead to severe economic hardships and loss of life in these domains. As fast as software spreads to automate many facets of our lives, it also grows in complexity. The complexity of software systems combined with the nature of the critical domains dependent …

Indoor Navigation Using Convolutional Neural Networks And Floor Plans, Ricky D. Anderson

Theses and Dissertations

The goal of this thesis is to evaluate a new indoor navigation technique by incorporating floor plans along with monocular camera images into a CNN as a potential means for identifying camera position. Building floor plans are widely available and provide potential information for localizing within the building. This work sets out to determine if a CNN can learn the architectural features of a floor plan and use that information to determine a location. In this work, a simulated indoor data set is created and used to train two CNNs. A classification CNN, which breaks up the floor plan into …

Comparison Of Machine Learning Techniques On Trust Detection Using Eeg, James R. Elkins

Theses and Dissertations

Trust is a pillar of society and is a fundamental aspect in every relationship. With the use of automated agents in todays workforce exponentially growing, being able to actively monitor an individuals trust level that is working with the automation is becoming increasingly more important. Humans often have miscalibrated trust in automation and therefore are prone to making costly mistakes. Since deciding to trust or distrust has been shown to correlate with specific brain activity, it is thought that there are EEG signals which are associated with this decision. Using both a human-human trust and a human-machine trust EEG dataset …

Solving The Quantum Layout Problem For Nisq-Era Quantum Computers Via Metaheuristic Algorithms, Brian D. Curran Jr.

Theses and Dissertations

In the noisy intermediate-scale quantum (NISQ)-era, quantum computers (QC) are highly prone to noise-related errors and suffer from limited connectivity between their physical qubits. Circuit transformations must be made to abstract circuits to address the noise and hardware constraints of NISQ-era devices. Such transformations introduce additional gates to the original circuit, thereby reducing the circuit's overall fidelity. To address the aforementioned constraints of NISQ-era QCs, dynamic remapping procedures permute logical qubits about physical qubits of the device to increase the fidelity of operations and make operations hardware-compliant. The quantum layout problem (QLP) is the problem of mapping logical qubits of …

Delayed Authentication System For Civilian Satellite, Sean M. Feschak

Theses and Dissertations

This thesis presents the feasibility of a Delayed Authentication System (DAS) for civilian satellite navigation (satnav) receivers. In satnav systems, encrypted signal components are transmitted synchronously with civilian components. Hence, the civilian signals can be authenticated by detecting the presence of encrypted signal components within the received signal. To authenticate, a reference station transmits estimated encrypted signal spreading code symbols processed using a high gain antenna. In this thesis, it is shown that a 1-meter diameter dish antenna is adequate to provide a high probability of successful authentication, thereby reducing overall system complexity and cost.