In this dissertation, we introduce a novel text representation method mainly used for text classification purpose. The presented representation method is initially based on a variety of closeness relationships between pairs of words in text passages within the entire corpus. This representation is then used as the basis for our multi-level lightweight ontological representation method (TOR-FUSE), in which documents are represented based on their contexts and the goal of the learning task. The method is unlike the traditional representation methods, in which all the documents are represented solely based on the constituent words of the documents, and are totally isolated from the goal that they are represented for. We believe choosing the correct granularity of representation features is an important aspect of text classification. Interpreting data in a more general dimensional space, with fewer dimensions, can convey more discriminative knowledge and decrease the level of learning perplexity. The multi-level model allows data interpretation in a more conceptual space, rather than only containing scattered words occurring in texts. It aims to perform the extraction of the knowledge tailored for the classification task by automatic creation of a lightweight ontological hierarchy of representations. In the last step, we will train a tailored ensemble learner over a stack of representations at different conceptual granularities. The final result is a mapping and a weighting of the targeted concept of the original learning task, over a stack of representations and granular conceptual elements of its different levels (hierarchical mapping instead of linear mapping over a vector). Finally the entire algorithm is applied to a variety of general text classification tasks, and the performance is evaluated in comparison with well-known algorithms.

In this dissertation, we introduce a novel text representation method mainly used for text classification purpose. The presented representation method is initially based on a variety of closeness relationships between pairs of words in text passages within the entire corpus. This representation is then used as the basis for our multi-level lightweight ontological representation method (TOR-FUSE), in which documents are represented based on their contexts and the goal of the learning task. The method is unlike the traditional representation methods, in which all the documents are represented solely based on the constituent words of the documents, and are totally isolated from the goal that they are represented for. We believe choosing the correct granularity of representation features is an important aspect of text classification. Interpreting data in a more general dimensional space, with fewer dimensions, can convey more discriminative knowledge and decrease the level of learning perplexity. The multi-level model allows data interpretation in a more conceptual space, rather than only containing scattered words occurring in texts. It aims to perform the extraction of the knowledge tailored for the classification task by automatic creation of a lightweight ontological hierarchy of representations. In the last step, we will train a tailored ensemble learner over a stack of representations at different conceptual granularities. The final result is a mapping and a weighting of the targeted concept of the original learning task, over a stack of representations and granular conceptual elements of its different levels (hierarchical mapping instead of linear mapping over a vector). Finally the entire algorithm is applied to a variety of general text classification tasks, and the performance is evaluated in comparison with well-known algorithms.

In this dissertation, we introduce a novel text representation method mainly used for text classification purpose. The presented representation method is initially based on a variety of closeness relationships between pairs of words in text passages within the entire corpus. This representation is then used as the basis for our multi-level lightweight ontological representation method (TOR-FUSE), in which documents are represented based on their contexts and the goal of the learning task. The method is unlike the traditional representation methods, in which all the documents are represented solely based on the constituent words of the documents, and are totally isolated from the goal that they are represented for. We believe choosing the correct granularity of representation features is an important aspect of text classification. Interpreting data in a more general dimensional space, with fewer dimensions, can convey more discriminative knowledge and decrease the level of learning perplexity. The multi-level model allows data interpretation in a more conceptual space, rather than only containing scattered words occurring in texts. It aims to perform the extraction of the knowledge tailored for the classification task by automatic creation of a lightweight ontological hierarchy of representations. In the last step, we will train a tailored ensemble learner over a stack of representations at different conceptual granularities. The final result is a mapping and a weighting of the targeted concept of the original learning task, over a stack of representations and granular conceptual elements of its different levels (hierarchical mapping instead of linear mapping over a vector). Finally the entire algorithm is applied to a variety of general text classification tasks, and the performance is evaluated in comparison with well-known algorithms.

Purpose – Artificial Intelligence has seen important growth in the digital area in recent years. Our aim is to explore possible guidelines that make use of AI advances to design good user experiences for digital products. Method – The proposed methods to gather the necessary qualitative data to support our claim involve open-ended interviews with UX/UI Designers working in the industry, in order to gain a deeper understanding of their thoughts and experiences. In addition, a literature review is conducted to identify the knowledge gap and build the base of our new theory. Findings – Our findings suggest a need to embrace new technological developments in favour of enhancing UX designers’ workflow. Additionally, basic AI and ML knowledge is needed to utilise these capabilities to their full potential. Indeed, a crucial area of impact where AI can augment a designer’s reach is personalization. Together with smart algorithms, designers may target their creations to specific user needs and demands. UX designers even have the opportunity for innovation as mundane tasks are automated by intelligent assistants, which broadens the possibility of acquiring further skills to enhance their work. One result, that is both innovative and unexpected, is the notion that AI and ML can augment a designer’s creativity by taking over mundane tasks, as well as, providing assistance with certain graphics and inputs. Implications – These results indicate that AI and ML may potentially impact the UX industry in a positive manner, as long as designers make use of the technology for the benefit of the user in true human-centred practice. Limitations – Nevertheless, our study presents its own unique limitations due to the scope and time frame of this dissertation, we are bound to the knowledge gathered from a small sample of professionals in Sweden. Presented guidelines are a suggestion based on our research and not a definitive workflow.

Artificiell intelligence (AI) har blivit vanligare att använda för att stödja människor i deras beslutsfattande. Maskininlärning (ML) är ett delområde inom AI som har börjat användas mer inom hälso-och sjukvården. Patientdata ökar inom vården och ett AI-system kan behandla denna ökade datamängd, vilket vidare kan utveckla ett beslutsstöd som hjälper läkarna. AI-tekniken blir vanligare att använda inom radiologin och specifikt inom mammografin som ett beslutsstöd. Användning av AI-teknik inom mammografin medför fördelar men det finns även utmaningar som inte har något med tekniken att göra.Icke-tekniska utmaningar är viktiga att se över för att generera en lyckad praxis. Studiens syfte var därför att undersöka icke-tekniska utmaningar vid användning av AI som ett beslutsstöd inom mammografi ur ett radiologiskt perspektiv. Radiologer med erfarenhet av mammografi intervjuades i syfte att öka kunskapen kring deras syn på användningen.Resultatet från studien identifierade och utvecklade de icke-tekniska utmaningarna utifrån temana: ansvar, mänskliga förmågor, acceptans, utbildning/kunskap och samarbete. Resultatet indikerade även på att inom dessa teman finns icke-tekniska utmaningar med tillhörande aspekter som är mer framträdande än andra. Studien ökar kunskaperna kring radiologers syn på användningen och bidrar till framtida forskning för samtliga berörda aktörer. Forskning kan ta hänsyn till dessa icke-tekniska utmaningar redan innan tekniken är implementerad i syfte att minska risken för komplikationer. / Artificial intelligence (AI) has become more commonly used to support people when making decisions. Machine learning (ML) is a sub-area of AI that has become more frequently used in health care. Patient data is increasing in healthcare and an AI system can help to process this increased amount of data, which further can develop a decision support that can help doctors. AI technology is becoming more common to use in radiology and specifically in mammography, as a decision support. The usage of AI technology in mammography has many benefits, but there are also challenges that are not connected to technology.Non-technical challenges are important to consider and review in order to generate a successful practice. The purpose of this thesis is therefore to review non-technical challenges when using AI as a decision support in mammography from a radiological perspective. Radiologists with experience in mammography were interviewed in order to increase knowledge about their views on the usage.The results identified and developed the non-technical challenges based on themes: responsibility, human abilities, acceptance, education/knowledge and collaboration. The study also found indications within these themes that there are non-technical challenges with associated aspects that are more prominent than others. This study emphasizes and increases the knowledge of radiologists views on the usage of AI and contributes to future research for all the actors involved. Future research can address these non-technical challenges even before the technology is implemented to reduce the risk of complications.

In today's society, companies are trying to find ways to utilize all the data they have, which considers valuable information and insights to make better decisions. This includes data used to keeping track of timber that flows between forest and industry. The growth of Artificial Intelligence (AI) and Machine Learning (ML) has enabled the development of ML modes to automate the measurements of timber on timber trucks, based on images. However, to improve the results there is a need to be able to get information from unlabeled images in order to decide weather and lighting conditions. The objective of this study is to perform an extensive for classifying unlabeled images in the categories, daylight, darkness, and snow on the load. A comparative study between partitional clustering and competitive learning is conducted to investigate which method gives the best results in terms of different clustering performance metrics. It also examines how dimensionality reduction affects the outcome. The algorithms K-means and Kohonen Self-Organizing Map (SOM) are selected for the clustering. Each model is investigated according to the number of clusters, size of dataset, clustering time, clustering performance, and manual samples from each cluster. The results indicate a noticeable clustering performance discrepancy between the algorithms concerning the number of clusters, dataset size, and manual samples. The use of dimensionality reduction led to shorter clustering time but slightly worse clustering performance. The evaluation results further show that the clustering time of Kohonen SOM is significantly higher than that of K-means.

Time setting in production operations is necessary to properly takt and balance the flow of assembly and logistics. Time setting activities is also crucial to achieve an optimised, healthy and ergonomic assembly and logistics operation. But time setting is seldom done on a detailed enough level before deployed on the shop floor which necessitates more work of the time setting to make it reflect the work carried out and fit it to the local production area. There is also a need to redo the time setting whenever a change to a process or product has occurred. Nowadays, the time setting is often performed using very manual methods with Predetermined Motion Time Systems (PMTS), sometimes with the aid of digital tools to replace pen and paper but work otherwise practically the same way it has since its inception in the first half of the 20th century. This is a process that require skill, experience and often much time, but is also monotonous and repetitive. To aid in the time setting process, and bring PMTS into Industry 5.0; a digitalised, smart tool is proposed where video can be used to feed a computer program to do the movement classification and time setting accurately and faster than current manual processes can achieve. However, the needs, challenges, and general function of such a system is not well researched in literature. This thesis thus delivers an analysis of current state for the time setting process at a large multinational truck manufacturer with production sites in Sweden and abroad, an overview of technologies for a digitalised, smart PMTS, and a conceptual framework for analysing production tasks using a digitalised, smart system. The framework is then partially implemented to showcase the usefulness of the system and how it would work in practice. / Korrekt tidssättning i produktion är nödvändigt för att takta, planera och balansera flödet i montering och logistik. Tidssättning är också avgörande för att uppnå en optimerad, hälsosam och ergonomisk monterings- och logistikverksamhet. Men tidssättningen görs sällan på en tillräckligt detaljerad nivå innan den används på verkstadsgolvet, vilket kräver mer arbete med tidssättningen för att den ska återspegla det utförda arbetet och anpassas till det lokala produktionsområdet. Det finns också ett behov av att göra om tidssättningen när en förändring av en process eller produkt har skett. Nuförtiden utförs tidssättningen ofta med väldigt manuella metoder med förutbestämda metod-rörelsesystem (PMTS), ibland med hjälp av digitala verktyg som ersätter penna och papper, men i övrigt fungerar det praktiskt taget på samma sätt som det har gjort sedan starten under första halvan av 1900-talet. Detta är en uppgift som kräver skicklighet, erfarenhet och ofta mycket tid, men som också är monoton och repetitiv. För att underlätta tidssättningsprocessen och ta förutbestämda metod-rörelsesystem in i Industri 5.0 föreslås nu ett digitaliserat, smart verktyg där video kan användas för att mata ett datorprogram som gör rörelseklassificeringen och tidssättningen mer exakt och snabbare än vad nuvarande manuella processer kan uppnå. De behov, utmaningar och den allmänna funktionen hos ett sådant system är dock inte väl undersökt i litteraturen utan kräver mer forskning. Detta examensarbete ger därför en analys av det nuvarande läget för tidssättningsprocessen hos en stor multinationell lastbilstillverkare med produktionsanläggningar i Sverige och utomlands, en översikt över tekniker för ett digitaliserat, smart PMTS och ett konceptuellt ramverk för analys av produktionsaktiviteter med hjälp av ett digitaliserat, smart system. Ramverket implementeras sedan delvis i en demonstrator för att visa hur ett sådant system kan se ut och fungera i praktiken.

M. Tech. (Department of Information Technology, Faculty of Applied and Computer Sciences), Vaal University of Technology. / This research was initiated because of a global increase in congestion on roads and the consequent increase in the rate of fatalities on both national and international roads. Annually, 1.3 million people are killed on the roads globally, and millions are injured. It was estimated that 2.4 million people will be killed in road traffic accidents annually by 2030, and in South Africa, over 14 000 deaths were reported in 2016. A study undertaken by the American Automobile Association Foundation for Traffic Safety (AAAFTS), established in 1947 to conduct research and address growing highway safety issues, found that motorcar accidents, on average, cost the United States $300 billion per annum. In the same vain, the World Health Organisation (WHO) asserted in their 2013 Global Status Safety Report on Road Safety that by 2020, traffic accidents would become the third leading cause of death globally. In this organisation’s 2015 report, South Africa was listed as having one of the highest road fatality rates in the world, averaging 27 out of 100 000 people. Cognisance of these statistics that describe wanton loss of life and serious economic implications, among other reasons, led to the development of autonomous vehicles (AVs), such as Google and Uber’s driverless taxis and Tesla’s autonomous vehicle. Companies have invested in self-driving prototypes, and they bolster this investment with continuous research to rectify imperfections in the technologies and to enable the implementation of AVs on conventional roads. This research aimed to address issues surrounding the systems communication concept, and focused on a novel method of the routing facet of AVs by exploring the mechanisms of the virtual system of packet switching and by applying these same principles to route autonomous vehicles. This implies that automated vehicles depart from a source address and arrive at a pre-determined destination address in a manner analogous to packet switching technology in computer networking, where a data packet is allotted a source and destination address as it traverses the Open Systems Interconnection (OSI) model for open system interconnection prior to dissemination through the network. This research aimed to develop an IoT model that reduces road congestion by means of a cost effective and reliable method of routing AVs and lessen dependency on vehicle-to-vehicle (V2V) communication with their heavy and costly sensor equipment and GPS, all of which under certain conditions malfunction. At the same time, as safety remains the foremost concern, the concept aimed to reduce the human factor to a considerable degree. The researcher demonstrated this by designing a computer-simulated Internet of Things (IoT) model of the concept. Experimental research in the form of a computer simulation was adopted as the most appropriate research approach. A prototype was developed containing the algorithms that simulated the theoretical model of IoT vehicular technology. The merits of the constructed prototype were analysed and discussed, and the results obtained from the implementation exercise were shared. Analysis was conducted to verify arguments on assumptions to clarify the theory, and the outcome of the research (an IoT model encompassing vehicular wireless technologies) shows how the basic concept of packet switching can be assimilated as an effective mechanism to route large-scale autonomous vehicles within the IoT milieu, culminating in an effective commuter operating system. Controlled routing will invariably save the traveller time, provide independence to those who cannot drive, and decrease the greenhouse effect, whilst the packet switching characteristic offers greater overall security. In addition, the implications of this research will require a workforce to supplement new growth opportunities.

<p>This thesis aims to design and develop the spatial adaptation approach through spatial transformers to improve the accuracy of human keypoint recognition models. We have studied different model types and design choices to gain an accuracy increase over models without spatial transformers and analyzed how spatial transformers increase the accuracy of predictions. A neural network called Widenet has been leveraged as a specialized network for providing the parameters for the spatial transformer. Further, we have evaluated methods to reduce the model parameters, as well as the strategy to enhance the learning supervision for further improving the performance of the model. Our experiments and results have shown that the proposed deep learning framework can effectively detect the human key points, compared with the baseline methods. Also, we have reduced the model size without significantly impacting the performance, and the enhanced supervision has improved the performance. This study is expected to greatly advance the deep learning of human key points and movement dynamics. </p>

The industrial production of semiconductor assemblies is subject to high requirements. As a result, several tests are needed in terms of component quality. In the long run, manual quality assurance (QA) is often connected with higher expenditures. Using a technique based on machine learning, some of these tests may be carried out automatically. Deep neural networks (NN) have shown to be very effective in a diverse range of computer vision applications. Especially convolutional neural networks (CNN), which belong to a subset of NN, are an effective tool for image classification. Deep NNs have the disadvantage of requiring a significant quantity of training data to reach excellent performance. When the dataset is too small a phenomenon known as overfitting can occur. Massive amounts of data cannot be supplied in certain contexts, such as the production of semiconductors. This is especially true given the relatively low number of rejected components in this field. In order to prevent overfitting, a variety of image augmentation methods may be used to the process of artificially creating training images. However, many of those methods cannot be used in certain fields due to their inapplicability. For this thesis, Infineon Technologies AG provided the images of a semiconductor component generated by an ultrasonic microscope. The images can be categorized as having a sufficient number of good and a minority of rejected components, with good components being defined as components that have been deemed to have passed quality control and rejected components being components that contain a defect and did not pass quality control. The accomplishment of the project, the efficacy with which it is carried out, and its level of quality may be dependent on a number of factors; however, selecting the appropriate tools is one of the most important of these factors because it enables significant time and resource savings while also producing the best results. We demonstrate a data augmentation graphical user interface (GUI) tool that has been widely used in the domain of image processing. Using this method, the dataset size has been increased while maintaining the accuracy-time trade-off and optimizing the robustness of deep learning models. The purpose of this work is to develop a user-friendly tool that incorporates traditional, advanced, and smart data augmentation, image processing, and machine learning (ML) approaches. More specifically, the technique mainly uses are zooming, rotation, flipping, cropping, GAN, fusion, histogram matching, autoencoder, image restoration, compression etc. This focuses on implementing and designing a MATLAB GUI for data augmentation and ML models. The thesis was carried out for the Infineon Technologies AG in order to address a challenge that all semiconductor industries experience. The key objective is not only to create an easy- to-use GUI, but also to ensure that its users do not need advanced technical experiences to operate it. This GUI may run on its own as a standalone application. Which may be implemented everywhere for the purposes of data augmentation and classification. The objective is to streamline the working process and make it easy to complete the Quality assurance job even for those who are not familiar with data augmentation, machine learning, or MATLAB. In addition, research will investigate the benefits of data augmentation and image processing, as well as the possibility that these factors might contribute to an improvement in the accuracy of AI models.