Search

Augmented Reality (AR) is a cutting-edge interactive technology. While Virtual Reality (VR) is based on completely virtual and immersive environments, AR superimposes virtual objects onto the real world. The value of AR has been demonstrated and applied within numerous industrial application areas due to its capability of providing interactive interfaces of visualized digital content. AR can provide functional tools that support users in undertaking domain-related tasks, especially facilitating them in data visualization and interaction by jointly augmenting physical space and user perception. Making effective use of the advantages of AR, especially the ability which augment human vision to help users perform different domain-related tasks is the central part of my PhD research. Industrial process tomography (IPT), as a non-intrusive and commonly-used imaging technique, has been effectively harnessed in many manufacturing components for inspections, monitoring, product quality control, and safety issues. IPT underpins and facilitates the extraction of qualitative and quantitative data regarding the related industrial processes, which is usually visualized in various ways for users to understand its nature, measure the critical process characteristics, and implement process control in a complete feedback network. The adoption of AR in benefiting IPT and its related fields is currently still scarce, resulting in a gap between AR technique and industrial applications. This thesis establishes a bridge between AR practitioners and IPT users by accomplishing four stages. First of these is a need-finding study of how IPT users can harness AR technique was developed. Second, a conceptualized AR framework, together with the implemented mobile AR application developed in an optical see-through (OST) head-mounted display (HMD) was proposed. Third, the complete approach for IPT users interacting with tomographic visualizations as well as the user study was investigated. Based on the shared technologies from industry, we propose and examine an AR approach for visual search tasks providing visual hints, audio hints, and gaze-assisted instant post-task feedback as the fourth stage. The target case was a book-searching task, in which we aimed to explore the effect of the hints and the feedback with two hypotheses: that both visual and audio hints can positively affect AR search tasks whilst the combination outperforms the individuals; that instant post-task feedback can positively affect AR search tasks. The proof-of-concept was demonstrated by an AR app in an HMD with a two-stage user evaluation. The first one was a pilot study (n=8) where the impact of the visual hint in benefiting search task performance was identified. The second was a comprehensive user study (n=96) consisting of two sub-studies, Study I (n=48) and Study II (n=48). Following quantitative and qualitative analysis, our results partially verified the first hypothesis and completely verified the second, enabling us to conclude that the synthesis of visual and audio hints conditionally improves AR search task efficiency when coupled with task feedback.

Software development is a messy process filled with an assortment of widely varying practices, procedures, and activities. Software refactoring is one of many such activities squeezed in to tight deadlines and frantic work schedules. Refactoring is often discussed as an isolated process, yet developers most often perform floss-refactoring where refactoring operations are performed along-side other development activities such as the addition of a new feature or to fix a bug. This has created a niche for automated tools to assist developers with the time-intensive tasks and the high cognitive cost associated with the practice of constant task switching. The various automated tools created to support refactoring currently rely on combinations of quality metrics and code anti-patterns, often called code smells or code anomalies, to find opportunities to improve the code. Yet, the tools do not use more task-specific knowledge to find opportunities to further improve both the refactoring recommendations and the associated tasks. Additionally, these tools frequently produce an overwhelming number of refactoring recommendations. The overwhelming nature of these recommendation options and competing quality objectives makes refactoring code quite challenging. To further add to this already onerous effort, these refactoring recommendations are presented to developers with little concern for how these refactorings are linked together or need to be applied to the code creating a truly herculean task. To address these challenges, we present the following contributions: (1) A simultasking, search-based algorithm that unifies the tasks of regression test case selection and software refactoring; (2) A theory, algorithm, and web-tool for defining and detecting ordering correctness dependencies between refactorings in a collection which we organize as graphs; and (3) A classifier for 3 architecture anti-patterns based on 16 code anomalies to explore the relationship between code-level and architecture-level issues.

Context: Automated Vehicle (AV) technology has evolved significantly in complexity and impact; it is expected to ultimately change urban transporta- tion. However, research shows that vehicle automation can only live up to this expectation if it is defined with human capabilities and limitations in mind. Therefore, it is necessary to bring human factors knowledge to AV developers. Objective: This thesis aims to empirically study how we can effectively bring the required human factors knowledge into large-scale agile AV develop- ment. The research goals are 1) to explore requirements engineering and human factors in agile AV development, 2) to investigate the problems of requirements engineering, human factors, and agile way of working in AV development, and 3) to demonstrate initial solutions to existing problems in agile AV development. Method: We conducted this research in close collaboration with industry, using different empirical methodologies to collect data—including interviews, workshops, and document analysis. To gain in-depth insights, we did a qualita- tive exploratory study to investigate the problem and used a design science approach to develop initial solution in several iterations. Findings and Conclusions: We found that applying human factors knowledge effectively is one of the key problem areas that need to be solved in agile development of artificial intelligence (AI)-intense systems. This motivated us to do an in-depth interview study on how to manage human factors knowl- edge during AV development. From our data, we derived a working definition of human factors for AV development, discovered the relevant properties of agile and human factors, and defined implications for agile ways of working, managing human factors knowledge, and managing requirements. The design science approach allowed us to identify challenges related to agile requirements engineering in three case companies in iterations. Based on these three case studies, we developed a solution strategy to resolve the RE challenges in agile AV development. Moreover, we derived building blocks and described guide- lines for the creation of a requirements strategy, which should describe how requirements are structured, how work is organized, and how RE is integrated into the agile work and feature flow. Future Outlook: In future work, I plan to define a concrete requirement strategy for human factors knowledge in large-scale agile AV development. It could help establishing clear communication channels and practices for incorporating explicit human factors knowledge into AI-based large-scale agile AV development.

Objective: To evaluate asynchronous digital visits (eVisits) with regard to digital communication, clinical decisionmaking,and subsequent care utilization in the digital era of primary care in Sweden.Methods: A mixed-methods approach was adopted across the various papers in the thesis, with all studiesevaluating the eVisit platform Flow in various clinical contexts.- Paper I was a comparative study of digital triage decisions when presented with automated patienthistory reports generated by the platform. Inter-rater reliability of triage decisions by majority vote in apanel of five physicians was compared to triage decisions by a machine learning model trained usingdata labelled by an expert primary care physician.- Paper II was a qualitative focus group study of nurse and physician experiences of digitalcommunication at three primary health care centers using the platform. Themes were generated usingqualitative content analysis as described by Graneheim and Lundman.- Papers III and IV were observational studies comparing office visits in the Skåne Region from Capio,a large private health care provider, to eVisit patients from Capio Go, a national eVisit service. Adultpatients with a chief complaint of sore throat, dysuria, or cough/common cold/influenza were recruited.eVisit patients were recruited prospectively digitally prior to their eVisit, while the office visit controlgroup was recruited retrospectively using letters. Paper III primarily compared antibiotic prescriptionrates per sore throat visit, while paper IV primarily compared subsequent physical health careutilization within two weeks for patients in the Skåne Region.Results: Interrater reliability was low (Cohen κ 0.17) between the panel majority vote and the machine learningmodel. Physicians and nurses experienced digitally filtered primary care, adjusting to a novel medium ofcommunication highlighting challenges in interpreting symptoms through text as well as alterations in practiceworkflow using asynchronous communication. Antibiotics prescription rate within three days was not higher aftereVisits compared to office visits (169/798 (21.2%) vs. 124/312 (39.7%) for sore throat, respectively; P

Traditionally, applications are executed without the notion of a computational deadline and often use all available system resources, which leads to higher energy consumption. User specification of Quality of Service (QoS) constraints, in terms of completion time and solution quality, opens up for allocation of just enough resources to an application to finish just in time and thereby save energy. Modern heterogeneous multiprocessor (HMP) platforms provide a set of configurable resources, including a frequency range of dynamic voltage frequency scaling (DVFS), one among a set processor types, and one or a plurality of processors of each type. They can be configured at run-time to open up new opportunities for resource management. This thesis presents techniques to reduce energy consumption under QoS constraints by allocating resources at run-time on heterogeneous multiprocessor platforms targeting sequential and parallel iterative and task-parallel applications. The proposed techniques rely on a progress-tracking framework that monitors and predicts how much time is left until the application finishes. Furthermore, the proposed framework enables the prediction of computation demand and performance requirements for future iterations or tasks. The first contribution of this thesis is a resource management technique, called SLOOP, targeting single-threaded applications. SLOOP allocates resources, i.e., processor type and DVFS, for each iteration to meet deadlines while using the prediction of computational demand and execution time. The second contribution of this thesis is a resource-management scheme, called SaC, for multi-threaded applications executing on HMPs, where resources also include the number of processors besides DVFS and processor type. SaC first chooses the most energy-efficient configuration that meets the deadline. The proposed technique collects execution-time slack over subsequent iterations to select a configuration that can save energy. The third contribution of this thesis is a resource manager, called Task-RM, for task-parallel applications executing on HMPs under QoS constraints. Task-RM exploits the variance in task execution times and imbalance between sibling tasks to allocate just enough resources in terms of DVFS and processor type. It uses an innovative off-line analysis to avoid redoing scheduling analysis at run-time. Finally, the fourth contribution is a scheme, called Approx-RM, that can exploit accuracy-energy trade-offs in approximate iterative applications. Approx-RM allocates an appropriate amount of resources while guaranteeing timing and solution quality specifications. Approx-RM first predicts the iteration count required to meet the quality target and then allocates enough resources on an HMP in terms of DVFS, processor type, and processor count to save energy while meeting a performance target.

Cilj diplomske naloge je bilo razviti grafični uporabniški vmesnik za vizualizacijo, procesiranje in analizo posnetkov elektrohisterograma (EHG). Razvili smo grafični uporabniški vmesnik EHGLab, prilagojen za delo s posnetki EHG v formatu WaveForm DataBase (WFDB). Vmesnik EHGLab omogoča prikaz signalov v časovnem in frekvenčnem prostoru. Za izločanje motenj v signalih smo načrtali Butterworthove digitalne filtrov z uporabo dvosmerne sheme filtriranja. Uporabniški vmesnik omogoča osnovno manipulacijo nad posnetki EHG, kot je postavljanje oznak in izrezovanje določenih delov posnetkov. Za ocenjevanje amplitudnih in močnostnih spektrov smo uporabili algoritem hitre diskretne Fourierjeve transformacije in avtoregresivno metodo. Vmesnik omogoča tudi analizo izračunanih močnostnih spektrov v izbranem frekvenčnem področju z iskanjem vrha spektra, frekvence vrha spektra in mediane frekvence. Vmesnik je razvit v okoljih NetBeans in Java. The goal of the thesis was to develop a graphic user interface for visualization, processing and analysis of electrohysterogram (EHG) records. We developed the graphic user interface EHGLab, which is adapted for work with the EHG records in the WaveForm DataBase (WFDB) format. The EHGLab interface allows display of the signals in time and frequency domain. To reject noise in signals we designed Butterworth digital filters using bidirectional filtering scheme. The user interface allows basic manipulation of the EHG records, like setting annotations and cutting selected parts of the records. For the estimation of the amplitude and power spectra we used the fast Fourier transformation algorithm and autoregressive method. The interface allows also analysis of the computed power spectra in the selected frequency band by seeking for the peak spectrum, the peak frequency of the spectrum, and the median frequency. The interface was developed in the NetBeans and Java environments.

A single gene can encode for different protein versions through a process called alternative splicing. Since proteins play major roles in cellular functions, aberrant splicing profiles can result in a variety of diseases, including cancers. Alternative splicing is determined by the gene's primary sequence and other regulatory factors such as RNA-binding protein levels. With these as input, we formulate the prediction of RNA splicing as a regression task and build a new training dataset (CAPD) to benchmark learned models. We propose discrete compositional energy network (DCEN) which leverages the hierarchical relationships between splice sites, junctions and transcripts to approach this task. In the case of alternative splicing prediction, DCEN models mRNA transcript probabilities through its constituent splice junctions' energy values. These transcript probabilities are subsequently mapped to relative abundance values of key nucleotides and trained with ground-truth experimental measurements. Through our experiments on CAPD, we show that DCEN outperforms baselines and ablation variants., ACM CHIL 2021 Camera-Ready

Digitalization is changing society, industry, and how business is done. For new companies that are more or less born digital, there is the opportunity to use and benefit from the capabilities offered by the new digital technologies, of which data-driven decision-making forms a crucial part. The manufacturing industry is facing the Fourth Industrial Revolution, but most manufacturing organizations are lagging behind in their digital transformation. This is due to the technical and organizational challenges they are experiencing. Based on this current state description and existing gap, the vision, aim, and research questions of this thesis are: Vision - future manufacturing organization to be driven by fact-based decision support based on data rather than of relying mainly on intuition and experience. Aim - to show manufacturing organizations the applicability of digital technologies in digitalizing manufacturing system data to support decision-making and how data sharing may be achieved. Research Question 1. How do manufacturing system lifecycle decisions influence the requirements of data collection towards interoperability? Research Question 2. What makes interoperability standardization applicable to sharing data in a manufacturing system’s lifecycle? This research is applied, addressing real-world problems in manufacturing. For this reason, the main objective is to solve the problem at hand, and data collection methods will be selected that can help address it. This can best be explained by taking a pragmatic worldview and using mixed methods approach that combines quantitative and qualitative methods. The research upon which this thesis is based draws on the results of three research projects involving the active participation of manufacturing companies. The data collection methods included experiments, interviews (focus group and semi-structured), technical development, literature review, and so on. The results are divided into three areas: 1) connected factory, 2) standard representation of machine model data, and 3) the digital twin in smart manufacturing. Connected factory addresses the question of how a mobile connectivity solution, 5G, may be used in a factory setting and demonstrates how the connectivity solution should be planned and how new data from a connected machine may support an operator in decision-making. The standard representation of machine model data demonstrates how an international standard may be used more widely to support the sharing and reuse of information. The digital twin in smart manufacturing investigates the reasons why there are so few real-world examples of this. The findings reveal that a manufacturing system’s lifecycle impacts data requirements, including a need for data accuracy in design, speed of data in operation (to allow operators to act upon events), and availability of historical data in maintenance (for finding root causes and planning). The volume of data was identified as important to all lifecycles. The applicability of standards was found to depend on: 1) the technology providers’ willingness to adapt standards, 2) enforcement by OEMs and larger actors further down a supply chain, 3) the development of standards that consider the user, and 4) when standards are required for infrastructure reasons. Based on the results and findings obtained, it may be stated that it is important to determine what actual manufacturing problem should be addressed by digital technology. There is a tendency to view this change from the perspective of what the digital technology might solve (a technology push), rather than setting aside the solution and focusing on what problem should be solved (a technology pull). This work also reveals the importance of the collaboration between industry and academia making progress in the digital transformation of manufacturing. Proofs-of-concept and demonstrators of how digital technologies might be used are also important tools in helping industry in how they can address a digital transformation.

The increasing connectivity is exposing safety-critical systems to cyberattacks that can cause real physical damage and jeopardize human lives. With billions of IoT devices added to the Internet every year, the cybersecurity landscape is drastically shifting from IT systems and networks to systems that comprise both cyber and physical components, commonly referred to as cyber-physical systems (CPS). The difficulty of applying classical IT security solutions in CPS environments has given rise to new security techniques known as process-aware defense mechanisms, which are designed to monitor and protect industrial processes supervised and controlled by cyber elements from sabotage attempts via cyberattacks. In this thesis, we critically examine the emerging CPS-driven cybersecurity landscape and investigate how process-aware defenses can contribute to the sustainability of highly connected cyber-physical systems by making them less susceptible to crippling cyberattacks. We introduce a novel data-driven model-free methodology for real-time monitoring of physical processes to detect and report suspicious behaviour before damage occurs. We show how our model-free approach is very lightweight, does not require detailed specifications, and is applicable in various CPS environments including IoT systems and networks. We further design, implement, evaluate, and deploy process-aware techniques, study their efficacy and applicability in real-world settings, and address their deployment challenges.

Automatic sleep stage mymargin classification is of great importance to measure sleep quality. In this paper, we propose a novel attention-based deep learning architecture called AttnSleep to classify sleep stages using single channel EEG signals. This architecture starts with the feature extraction module based on multi-resolution convolutional neural network (MRCNN) and adaptive feature recalibration (AFR). The MRCNN can extract low and high frequency features and the AFR is able to improve the quality of the extracted features by modeling the inter-dependencies between the features. The second module is the temporal context encoder (TCE) that leverages a multi-head attention mechanism to capture the temporal dependencies among the extracted features. Particularly, the multi-head attention deploys causal convolutions to model the temporal relations in the input features. We evaluate the performance of our proposed AttnSleep model using three public datasets. The results show that our AttnSleep outperforms state-of-the-art techniques in terms of different evaluation metrics. Our source codes, experimental data, and supplementary materials are available at https://github.com/emadeldeen24/AttnSleep. Agency for Science, Technology and Research (A*STAR) Published version The work of Emadeldeen Eldele was supported by A*STAR SINGA Scholarship.

A methodology is proposed in this paper allowing the classification of anomalies and subsequently their possible localization in nuclear reactor cores during operation. The method relies on the monitoring of the neutron noise recorded by in-core neutron detectors located at very few discrete locations throughout the core. In order to unfold from the detectors readings the necessary information, a 3-dimensional Convolutional Neural Network is used, with the training and validation of the network based on simulated data. In the reported work, the approach was also tested on simulated data. The simulations were carried out in the frequency domain using the CORE SIM+ diffusion-based two-group core simulator. The different scenarios correspond to the following cases: a generic “absorber of variable strength”, axially travelling perturbations at the velocity of the coolant flow (due to e.g. fluctuations of the coolant temperature at the inlet of the core), fuel assembly vibrations, control rod vibrations, and core barrel vibrations. In all those cases, various frequencies were considered and, when relevant, different locations of the perturbations and different vibration modes were taken into account. The machine learning approach was able to correctly identify the different scenarios with a maximum error of 0.11%. Moreover, the error in localizing anomalies had a mean squared error of 0.3072 in mesh size, corresponding to less than 4 cm. The proposed methodology was also demonstrated to be insensitive to parasitic noise and will be tested on actual plant data in the near future.

Although parallel programming languages have existed for decades, (scientific) parallel programming is still dominated by Fortran and C/C++ augmented with parallel programming frameworks, e.g., MPI, OpenMP, OpenCL and CUDA. We perform a comparative study of Chapel and Julia, two languages quite different from one another as well as from Fortran and C, in regard to parallel programming on distributed and shared memory computers. The study is carried out using test cases that expose the need for different approaches to parallel programming. Test cases are implemented in Chapel and Julia, and in C augmented with MPI and OpenMP. It is shown that both languages, Chapel and Julia, represent a viable alternative to Fortran and C/C++ augmented with parallel programming frameworks: the programmer's efficiency is considerably improved while the speed of programs is not significantly affected.

We first describe elliptic curves and their arithmetic. Then we introduce the discrete logarithm problem and its implications in cryptographic protocols, such as the Diffie-Helman key agreement and the signatures DSA and ECDSA. The most important general attacks on the discrete logarithm problem are also studied, with an emphasis on those that work in the group of points on an elliptic curve, i.e., the giant step-baby step algorithm and the Pollard's rho algorithm together with the Floyd's algorithm for cycle detection. Our application for interactive representation of elliptic curves and their arithmetic is also described. It is divided into two smaller applications. The first focuses on elliptic curves over real numbers and the second one on prime fields.

To work with textual data, machine learning algorithms, in particular, neural networks, require word embeddings – vector representations of words in high-dimensional space. There are languages with a small amount of available resources. Exploiting the knowledge from the well-resourced languages for under-resourced languages is possible with cross-lingual embeddings by aligning the embeddings of one language with the vector space of another language. Existing methods for aligning embeddings are intended for context-independent embeddings, where every word has one representation. We propose a method, based on a dictionary and a parallel corpus aligns contextual embeddings, which capture more information about the context in which words appear. The proposed method requires a small amount of bilingual data, which is available for many language pairs. We empirically show that the proposed method outperforms the baseline obtained by alignment of context-independent embeddings. Da bi algoritmi strojnega učenja, še posebej nevronske mreže, delali z besedilnimi podatki, potrebujejo vložitve besed – predstavitev besed v visokodimenzionalnem vektorskem prostoru. Za nekatere jezike je na voljo le majhna količina jezikovnih virov. Zanje je pomembno izkoriščanje znanja iz tehnološko bolj razvitih jezikov, kar omogočajo medjezikovne vložitve. Te vektorski prostor besed enega jezika preslikajo v vektorski prostor drugega jezika. Obstoječe metode za poravnavo vektorskih vložitev so namenjene kontekstno neodvisnim vložitvam, kjer ima vsaka beseda le eno predstavitev. Obstoječe metode za poravnavo vektorskih vložitev so namenjene kontekstno neodvisnim vložitvam, kjer ima vsaka beseda le eno predstavitev. Predstavljamo pristop, ki na podlagi slovarja in paralelnega korpusa poravna kontekstne vložitve, ki zajemajo več informacij o kontekstu, v katerem so besede uporabljene. Predlagani pristop zahteva majhno količino dvojezičnih virov, ki so na voljo za mnogo parov jezikov. Empirično pokažemo, da je predlagani pristop opazno boljši od izhodiščnega, s katerim poravnavamo kontekstno neodvisne vložitve.

The diploma work is designed as a web portal that was built using standard web technologies with emphasis on the Laravel and Vue.js frameworks. The review of the field showed that the portal, which includes all the required goals of the thesis, is not yet available. The web portal displays news, external links, contacts and areas in the third focus area, and in the fourth focus area of the development and innovation partnership SRIP HRANA. It is also possible to review companies and research institutions. It is used to find potential opportunities for cooperation, leading to innovative solutions which increase productivity in the agro-food sector. Before the creation of the web portal, certain data were collected, and the administration page allows for these data to be added, edited or deleted. The administrative and operational functions of the application are described, and presented with screenshots illustrating potential options for upgrades and practical use.

One of the problems road holders are facing is maintaining a record of road's surface quality. They acquire a vast amount of image data and then assess the surface quality by manually inspecting those images, which is time consuming and often inconsistent. In this work we show how to tackle a similar problem of automatic recognition of road surface type. To solve this problem we use the artificial neural network for classification tasks based on ResNet-50 architecture. To boost it's performance we use the information of the road's position in the input image which is obtained with U-Net neural network for semantic segmentation. In case of segmentation we show how to emphasise pixels located near road's edges and focus the network's attention during training to the parts where errors are most frequent. We also consider coarsely annotated images and show how we can use unlabelled pixels assigning them lower weights during the training process. We compare two attention mechanisms for neural networks used for classification tasks. The first mechanism masks input images with zero values where segmentation network detects background. The second mechanism is based on extending the input image with an output of U-Net. We show that by using the second approach F1 score evaluated on the test dataset improves from 0.947 to 0.971.

The thesis presents the problem of designing and implementation of business processes using specific approach. We presented the whole path from defining the process to the final implementation and execution. The business process is defined by a graphic notation, since the visual presentation of data brings important advantages. The selected notation is used to model the process and convert it to source code as automatically as possible. Where automatic conversion was not possible, we used module MSM from Boost Library. We presented a general purpose model language UML, with the emphasis on activity diagram and state diagram. The BPMN standard for modeling business processes was introduced as well. BPMN has several common points with UML activity diagram. After the comparison we chose to use UML activity diagram for further purposes. To demonstrate the design and implementation of business processes, we defined the business process of the input invoice liquidation and modeled it using activity diagram. In addition, we used class diagram to define objects and generated the C++ code for them. With the help of the MSM module which is based on state machines, we implemented the input invoice liquidation. In this part the module is presented as well. The main thesis result is a test application which simulates the inputs of different users and leads the whole process of the input invoice liquidation. Finally the analysis of obtained result is presented in the final part of the thesis. The main thesis result is a test application which simulates the inputs of different users and leads the whole process of the input invoice liquidation. Finally the analysis of obtained result is presented in the final part of the thesis.

Reconstruction of geometry from RGB images is one of the classic computer vision problems. The quality of the produced 3D model heavily depends on the input images. Manual image acquisition can be a lengthy process with which we want to attain the desired accuracy and completeness of the model. A user that has no feedback about the suitability of the images during the acquisition process can have difficulties complying with assumptions of the algorithms, which can result in an unsuccessful reconstruction. Additional image acquisition can be expensive or even impossible. In this work, we focus on development of a system and software that support the entire reconstruction process. The user gets online information about the adequacy of every captured image and an estimate of quality for the current 3D model. We also present a novel method for next best view planning, that systematically improves the quality of reconstruction. The method is based on a new quality measure. We show that the linear correlation coefficient between our measure and accuracy is better than that of the existing measure. We also show that the reconstruction obtained by the next best view planning is comparable and in some cases better than reconstruction with evenly spaced camera configuration in the shape of a hemisphere.

Bluetooth Low Energy is one of the most popular technologies used by small electronic devices for wireless communication. Because it operates in an overcrowded frequency band it is exposed to a lot of interferences. This thesis focuses on the influence of Wi-Fi interferences on the communication of BLE devices. With practical experiments, we want to show different ways of countering interferences and analyze how successful BLE communication is. We found out, that interferences have an effect on advertising channels only in close proximity. When a BLE communication is established, Wi-Fi traffic on one Wi-Fi channel has little effect on the success rate of packets, as BLE uses a frequency hopping method, which almost completely negates any effect. When there is Wi-Fi traffic on the whole BLE frequency band only two channels remain, which are always the last two data channels.

Understanding the sense of discourse relations that appear between segments of text is essential to truly comprehend any natural language text. Several automated approaches have been suggested, but all rely on external resources, linguistic feature engineering, and their processing pipelines are built from substantially different models. Instead of designing a system specifically for a given language and task, we pursue a language-independent approach for sense classification of shallow discourse relations. In this dissertation we first present our focused recurrent neural networks (focused RNNs) layer, the first multi-dimensional RNN-attention mechanism for constructing sentence/argument embeddings. It consists of a filtering RNN with a filtering/gating mechanism that enables downstream RNNs to focus on different aspects of each argument of a discourse relation and project it into several embedding subspaces. On top of the proposed mechanism we build our FR system, a novel method for sense classification of shallow discourse relations. In contrast to existing systems, the FR system consists of a single end-to-end trainable model for handling all types and specific situations of discourse relations, requires no feature engineering or external resources, can be used almost out-of-the-box on any language or set of sense labels, and can be applied at the word and character level representation. We evaluate the proposed FR system using the official datasets and methodology of CoNLL 2016 Shared Task. It does not fall a lot behind state-of-the-art performance on English, but it outperforms other systems without a focused RNNs layer by 8% on the Chinese dataset. Afterwards we perform a detailed analysis on both languages. Razumevanje smisla diskurznih relacij, ki nastopajo med segmenti besedila, je ključnega pomena za razumevanje kateregakoli besedila v naravnem jeziku. Številni avtomatizirani pristopi so že bili predlagani, vendar so vsi odvisni od zunanjih virov, ročno-idelanih značilk in njihovi cevovodi za procesiranje so izdelani iz bistveno različnih modelov. Namesto izdelave sistema specializiranega za dani jezik in nalogo, mi stremimo k jezikovno-neodvisnemu pristopu za klasifikacijo smisla v plitkih diskurznih relacijah. V pričujoči disertaciji najprej predstavimo naše osredotočene rekurentne nevronske mreže (focused RNNs), ki predstavljajo prvi več-dimenzionalni RNN-pozornostni mehanizem za izdelavo vložitev stavkov/argumentov. Sestavljen je iz filtrirnega RNN z mehanizmom za filtriranje/usmerjanje, ki omogoča sledečim RNN-jem, da se osredotočijo na različne vidike vsakega argumenta diskuzne relacije in ga projecirajo v več vložitvenih podprostorov. Omenjeni mehanizem uporabimo v našem sistemu FR system, ki predstavlja novo metodo za klasifikacijo smisla v plitkih diskurznih relacijah. V nasprotju z obstoječimi sistemi je FR system sestavljen iz enega modela, ki ga je mogoče celostno učiti od začetka-do-kraja, obravnava vse vrste in specifične situacije v diskurznih relacijah, ne potrebuje ročno-izdelanih značilk ali zunanjih virov, se lahko skorajda brez sprememb uporabi na kateremkoli jeziku ali oznakah smisla, in se lahko uporablja tako na ravni besed kot na ravni znakov. Predlagani FR system smo ovrednotili na uradnih podatkovnih zbirkah in po metodologiji izziva CoNLL 2016 Shared Task. Ne zaostaja veliko za najuspešnejšimi sistemi na angleškem jeziku, vendar presega ostale sisteme brez focused RNNs plasti za 8% na kitajski podatkovni zbirki. Nato smo izvedli natančnejšo analizo na obeh jezikih.

The idea for this thesis is the development of a mobile application for Android devices that lets people share live music between two devices. As a first step, I had to solve how to find and play all the audio files located in the device. The next step was to connect two devices using Bluetooth and Wi-Fi Direct technologies and find a way to be able to listen the same song on both devices at the same time. Lastly, I worked on getting the geolocation from the device, so I could have a record of the travelled paths and distances.

In this thesis, an application with a graphical user interface for trading on crypto exchanges was created. This application has all the basic functionalities offered by today's trading platforms. These functionalities include: trading, reviewing user history, reviewing and managing user status. The application is written in the C# programming language using the .NET framework and currently supports trading on three different crypto exchanges: ForkDelta, KuCoin, and HitBTC. These stock exchanges have been added to the application as libraries and were also implemented in the framework of the thesis. In addition to the API of crypto exchanges, the application also uses other software interfaces: CoinMarketCap (for access to the average crypto prices on the market), Infura.io (for connection to the Ethereum blockchain, which we need to interface with the ForkDelta stock exchange). The application is written in a modular approach that allows the addition of the support for other crypto exchanges in the future.

This diploma compares React Native with native tools for developing mobile applications from the perspective of both the developer and user. Factors that impact project duration, quality of code and functionality of the final product like number of tools, their capabilities and simplicity of use have been considered for the development aspect. Functionality, user experience and responsiveness were assessed from the user's perspective, however they are mostly unnoticeable or non-existent due to the maturity of React Native. The biggest difference is felt by developers because of different programming languages used (Java for native and Javascript for React Native applications), however it does depend on the type of project and development team.

Application translators often face a problem of not finding an appropriate translation for a specific word or are unable to translate it, as it has multiple interpretations. These problems arise when a translator is unaware of the context the word is used in. So, the common practice for the developers is to export the words into a list, which then gets send to the translators. When the words are translated, they are then sent back to the developers, whose task it is to import them into their application, with which the application development process is greatly slowed down. If the translators come across the same problems of not knowing which is the appropriate translation, they are forced to ask for additional information. In smaller information systems the context is easily identified; whereas the more complicated require significantly more effort. In our attempt at fixing this issue We will develop a solution, which will enable translators to translate text based on the context of the information systems website. Translator will be required to enter a link to the website, from where our program will extract the words it needs to translate. Since our application will offer a direct translation, without the intermediate of developers, which would normally be tasked with inputting the information into a websites database, the whole translation process will improve in speed.

Catastrophic forgetting is phenomenon when an artificial neural network immediately and almost completely forgets previously learned tasks when trained incrementally on new ones. It is a well-known problem and although there are many approaches to alleviating it, none of them solves it completely. We experimentally check for main causes of catastrophic forgetting. Analysis is performed on a deep convolutional neural network for image classification. Results are interpreted by confusion matrices and classification accuracy graphs, we also visualize changes of weights and biases of network. Analytical findings serve as a basis for designing different approaches to updating network parameters, aiming to prevent or alleviate catastrophic forgetting. We also evaluate effects of availability of Oracle, capable of determining subset of all possible classes for classification, when using the network. We implement one of existing approaches to preventing catastrophic forgetting and adapt it to work without Oracle. Findings, presented in thesis serve as a starting point for design of new approaches aimed at preventing catastrophic forgetting.

Trends in modern industry strive to the rapid product development and their improvement. They often require traceability and quality assurance. Labelling supports the connection of physical products with the information saved in digital form. In most cases it does not bring additional value to products themselves, but it serves only for optimization and support to other existing processes. The main goal of this master thesis is an improvement of the existing printing process with the implementation of two cloud extensions. First, it supports integration of the labelling system with other systems and, second, it supports location independent label printing. The existing label management system with the implementation of those extensions provides added value, because it supports the integration of the printing process with other processes and supports their optimization.

We used a computer model containing a flock of fish and a predator to evaluate the effect of energy on predator speed and success of various attack tactics. We also looked into predator confusability and tried to determine whether fish form flocks as a form of a defense mechanism. Results were obtained using a genetic algorithm. We simulated natural evolution of predators and prey, and came to conclusion that the best attack tactic is highly dependant on predator's method of confusability. Moreover, we showed flocking increases the chance of individual survival in certain attack tactics, while it decreases the chance of individual survival in others.

This thesis presents issues with compatibility of implementation and usage of different Business Process Management Software. Some of those issues are longer developing cycles and an increased cost of the product. To overcome these issues, I developed a microservice, which works as a wrapper around BPMS and hides the specifics of a particular BPMS implementation and allows the user to be agnostic about the implementation. The microservice communicates through REST APIs with BPMS and aggregates their calls into unified calls, making it easier to develop and deliver a more affordable product.

This Bachelor’s Thesis describes the process of collecting and visualizing data from different sources. There are three different data sources. The first source is from the Statistical office of Slovenia where there is data about the number of baby names occurring from 1992 to 2017. The second sourse is the IMDb database, which has data about actors and movies. The third data source is the free Wikipedia encyclopedia , which holds interesting data about names. To be able to merge all the datasources requires a great range of frameworks. For importing the data, the programming language Python is used. For data storage about the number of babynames, the unrelation database Elasticsearch is used. For the exchange of data which is stored on the inter- net or on local machine servers, either Python or Node.js. are implemented. In addition, the basic web technologies JavaScript and D3.js are the main tools for data visualization.

My work describes functionalities of the SharePoint platform and how the functionality can be extended by additional applications like K2 Blackpearl (Business Process Management, forms designer) and ABBYY Flexi Caputre (OCR, capturing and processing paper documents) to design complete business solutions for managing and processing documents and other processes. Combination of all three is extremely strong, but easy to use tool. First application described in my work is pure SharePoint and basic tools that belong with it. Second application is also based on SharePoint, but is extended with K2, which significantly simplifies and speeds up application developments for business needs.

Purpose of this bachelor thesis is to present the process of software development with emphasis on software testing in one of the biggest Slovenian IT company and to do an objective evaluation based on the CMM model. We also present basic functionalities of the software testing tool SoapUI, which we used in the company. Throughout the analysis of the processes of software development and software testing, we find that the company is at the first, or initial level of the CMM model. We also present a few practical test cases with SoapUI tool and discover limitations in regard to its usage in the company's testing process. Based on these findings we then propose solutions, which can be directly implemented by the company. Key solutions that we propose for the company are to divide the analytic/tester role and introduce a new team for the sole purpose of testing computer software in the company, introduction of traineeship programmes for new employees and active tracking and logging of errors which are detected throughout the software development process.

Sokoban is a puzzle game that offers computationally complex problems that are very difficult to solve. There are many algorithms that can be used to solve these problems, but none is successful in solving all given, even relatively simple problems. Therefore, the game Sokoban is an interesting platform both for the improvement of the search algorithms and for a better understanding. In this thesis we propose an application that can be used both as a tool to support the developers of these algorithms, or as a means to analyze and teach their functioning. We present the development process of the application, test its functionality on practical Sokoban puzzles and show that it works correctly and is a relevant tool.

Resources in the information system are not unlimited. Therefore, we should monitor resource usage as we monitor the operation of the information system. Docker container as a part of an information system is our main monitoring objective in this thesis. A set of tools for monitoring Docker containers is presented: from a simple, in-built tool, to more complex SaaS service cloud monitoring tools. The Kubernetes platform monitoring tools are also presented. A practical example illustrates the use of some of the presented tools. Based on the comparative analysis, proposals are made for the use of proper monitoring tool in the aspect of the complexity and diversity of monitored elements and additional guidance that can be helpful in deciding which monitoring tool to use.