Astronomy has a data de-noising problem. The quantity of data produced by astronomical instruments is immense, and a wide variety of noise is present in this data including artifacts. Many types of this noise are not easily filtered using traditional handwritten algorithms. Deep learning techniques present a potential solution to the identification and filtering of these more difficult types of noise. In this thesis, deep learning approaches to two astronomical data de-noising steps are attempted and evaluated. Pre-existing simulation tools are utilized to generate a high-quality training dataset for deep neural network models. These models are then tested on real-world data. One set of models masks diffraction spikes from bright stars in James Webb Space Telescope data. A second set of models identifies and masks regions of the sky that would interfere with sky surface brightness measurements. The results obtained indicate that many such astronomical data de-noising and analysis problems can use this approach of simulating a high-quality training dataset and then utilizing a deep learning model trained on that dataset.

Open Information Extraction (OIE) is a subset of Natural Language Processing (NLP) that constitutes the processing of natural language into structured and machine-readable data. This thesis uses data in Resource Description Framework (RDF) triple format that comprises of a subject, predicate, and object. The extraction of RDF triples from natural language is an essential step towards importing data into web ontologies as part of the linked open data cloud on the Semantic web. There have been a number of related techniques for extraction of triples from plain natural language text including but not limited to ClausIE, OLLIE, Reverb, and DeepEx. This proposed study aims to reduce the dependency on conventional machine learning models since they require training datasets, and the models are not easily customizable or explainable. By leveraging a context-free grammar (CFG) based model, this thesis aims to address some of these issues while minimizing the trade-offs on performance and accuracy. Furthermore, a deep-dive is conducted to analyze the strengths and limitations of the proposed approach.

There exists extensive research on the use of twisty puzzles, such as the Rubik's Cube, in educational contexts to assist in developing critical thinking skills and in teaching abstract concepts, such as group theory. However, the existing research does not consider the use of twisty puzzles in developing language proficiency. Furthermore, there remain methodological issues in integrating standard twisty puzzles into a class curriculum due to the ease with which erroneous cube twists occur, leading to a puzzle scramble that deviates from the intended teaching goal. To address these issues, an extensive examination of the "smart cube" market took place in order to determine whether a device that virtualizes twisty puzzles while maintaining the intuitive tactility of manipulating such puzzles can be employed both to fill the language education void and to mitigate the potential frustration experienced by students who unintentionally scramble a puzzle due to executing the wrong moves. This examination revealed the presence of Bluetooth smart cubes, which are capable of interfacing with a companion web or mobile application that visualizes and reacts to puzzle manipulations. This examination also revealed the presence of a device called the WOWCube, which is a 2x2x2 smart cube entertainment system that has 24 Liquid Crystal Display (LCD) screens, one for each face's square, enabling better integration of the application with the puzzle hardware. Developing applications both for the Bluetooth smart cube using React Native and for the WOWCube demonstrated the higher feasibility of developing with the WOWCube due to its streamlined development kit as well as its ability to tie the application to the device hardware, enhancing the tactile immersion of the players with the application itself. Using the WOWCube, a word puzzle game featuring three game modes was implemented to assist in teaching players English vocabulary. Due to its incorporation of features that enable dynamic puzzle generation and resetting, players who participated in a user survey found that the game was compelling and that it exercised their critical thinking skills. This demonstrates the feasibility of smart cube applications in both critical thinking and language skills.

Traditional sports coaching involves face-to-face instructions with athletes or playingback 2D videos of athletes’ training. However, if the coach is not in the same area as the athlete, then the coach will not be able to see the athlete’s full body and thus cannot give precise guidance to the athlete, limiting the athlete’s improvement. To address these challenges, this paper proposes Augmented Coach, an augmented reality platform where coaches can view, manipulate and comment on athletes’ movement volumetric video data remotely via the network. In particular, this work includes a). Capturing the athlete’s movement video data with Kinects and converting it into point cloud format b). Transmitting the point cloud data to the coach’s Oculus headset via 5G or wireless network c). Coach’s commenting on the athlete’s joints. In addition, the evaluation of Augmented Coach includes an assessment of its performance from five metrics via the wireless network and 5G network environment, but also from the coaches’ and athletes’ experience of using it. The result shows that Augmented Coach enables coaches to instruct athletes from a distance and provide effective feedback for correcting athletes’ motions under the network.

This paper introduces Zenith, a statically typed, functional programming language that compiles to Lua modules. The goal of Zenith is to be used in tandem with Lua, as a secondary language, in which Lua developers can transition potentially unsound programs into Zenith instead. Here developers will be ensured a set of guarantees during compile time, which are provided through Zenith’s language design and type system. This paper formulates the reasoning behind the design choices in Zenith, based on prior work. This paper also provides a basic understanding and intuitions on the Hindley-Milner type system used in Zenith, and the functional programming data types used to encode unsound functions. With these ideas combined, the paper concludes on how Zenith can provide soundness and runtime safety as a language, and how Zenith may be used with Lua to create safe systems.

Recommendation systems provide recommendations based on user behavior andcontent data. User behavior and content data are fed to machine learning algorithms to train them and give recommendations to the users. These algorithms need a large amount of data for a reasonable conversion rate. But for small applications, the available amount of data is minimal, leading to high recommendation aberrations. Also, when an existing large scaled application with a high amount of available data uses a new recommendation system, it requires some time and testing to decide which recommendation algorithm is best suited to get higher conversion rates. This learning curve costs highly when the user base and data size are significantly high. In this thesis, A/B testing is used with manual intervention in the decision-making of recommendation systems. To understand the effectiveness of the recommendations, user interaction data is compared to compare experiences. Based on the comparisons, the experiments conclude the effectiveness of A/B testing for the recommendation system.

Since the early 2000s the Rubik’s Cube has seen growing usage at speedsolving competitions and as an effective tool to teach Science, Technology, Engineering, Mathematics (STEM) topics at hundreds of schools and universities across the world. Recently, cube manufacturers have begun embedding sensors to enable digital face tracking. The live feedback from these so called “smartcubes” enables a new wave of immersive solution tutorials and interactive educational games using the cube as a controller. Existing smartcube software has several limitations. Manufacturers’ applications support only a narrow set of puzzle form factors and application platforms, fragmenting the ecosystem. Most apps require an active internet connection for key features, limiting where users can practice with a smartcube. Finally, existing applications focus on a single 3x3x3connection, losing opportunities afforded by new form factors. This research demonstrates an open-source smartcube application which mitigates these limitations. Particular attention is given to creating an Application Programming Interface (API) for smartcube communication and building representative solve analysis tools. These innovations have included successful negotiations to re-license existing open-source Rubik’sCube software projects to support deployment on multiple platforms, particularly iOS. The resulting application supports smartcubes from three manufacturers, runs on two platforms (Android and iOS), functions entirely offline after an initial download of remote assets, demonstrates concurrent connections with up to six smartcubes, and supports all current and anticipated smartcube form factors. These foundational elements can accelerate future efforts to build smartcube applications, including automated performance feedback systems and personalized gamification of learning experiences. Such advances will hopefully enhance the Rubik’s Cube’s value both as a competitive toy and as a pedagogical tool in educational institutions worldwide.

Frontend development often involves the repetitive and time-consuming task of transforming a Graphical User interface (GUI) design into Frontend Code. The GUI design could either be an image or a design created on tools like Figma, Sketch, etc. This process can be particularly challenging when the website designs are experimental and undergo multiple iterations before the final version gets deployed. In such cases, developers work with the designers to make continuous changes and improve the look and feel of the website. This can lead to a lot of reworks and a poorly managed codebase that requires significant developer resources. To tackle this problem, researchers are exploring ways to automate the process of transforming image designs into functional websites instantly. This thesis explores the use of machine learning, specifically Recurrent Neural networks (RNN) to generate an intermediate code from an image design and then compile it into a React web frontend code. By utilizing this approach, designers can essentially transform an image design into a functional website, granting them creative freedom and the ability to present working prototypes to stockholders in real-time. To overcome the limitations of existing publicly available datasets, the thesis places significant emphasis on generating synthetic datasets. As part of this effort, the research proposes a novel method to double the size of the pix2code [2] dataset by incorporating additional complex HTML elements such as login forms, carousels, and cards. This approach has the potential to enhance the quality and diversity of training data available for machine learning models. Overall, the proposed approach offers a promising solution to the repetitive and time-consuming task of transforming GUI designs into frontend code.

SLAM (Simultaneous Localization and Mapping) is a problem that has existed for a long time in robotics and autonomous navigation. The objective of SLAM is for a robot to simultaneously figure out its position in space and map its environment. SLAM is especially useful and mandatory for robots that want to navigate autonomously. The description might make it seem like a chicken and egg problem, but numerous methods have been proposed to tackle SLAM. Before the rise in the popularity of deep learning and AI (Artificial Intelligence), most existing algorithms involved traditional hard-coded algorithms that would receive and process sensor information and convert it into some solvable sensor-agnostic problem. The challenge for these sorts of methods is having to tackle dynamic environments. The more variety in the environment, the poorer the results. Also due to the increase in computational power and the capability of deep learning-based image processing, visual SLAM has become extremely viable and maybe even preferable to traditional SLAM algorithms. In this research, a deep learning-based solution to the SLAM problem is proposed, specifically monocular visual SLAM which is solving the problem of SLAM purely with a singular camera as the input, and the model is tested on the KITTI (Karlsruhe Institute of Technology & Toyota Technological Institute) odometry dataset.