In the banking world trades of securities are finalized every day, on behalf of the banks themselves or of their clients. When the trades have been booked by the front office the confirmations sent by the counterparty have to be checked and connected to the correct trade by hand, posing the question whether this process could not be automated using machine learning techniques. There is no straightforward solution to this problem since the confirmations differ between counterparties, and can contain different enriched information or even be in different formats. This thesis addresses the problem of matching trades with their corresponding confirmations via deep learning methods. A model is trained using contrastive learning methods on generated pairs of trades and confirmations, with the goal of matching the pairs in the latent space by using nearest neighbor classification. Accuracy is measured by dividing the correctly classified samples by the total number of samples in a testing batch. The model achieves an accuracy as high as 97.8% over 100 trade-confirmation samples with a 30-dimensional latent space, and it is shown that similar contrastive methods can indeed be used in order to solve this problem. / Banker handlar varje dag med värdpapper av olika slag, antingen för sin egen vinning eller för sina kunders. När en affär har blivit beslutad mellan två parter så bokförs denna i bägge parternas interna system. En konfirmation kommer sedan skickas från den andra parten som manuellt måste paras ihop med affären vilket väcker frågan om huruvida detta inte kan automatiseras med hjälp av maskininlärning. Det finns inte en uppenbar lösning på detta problemet då konfirmationsmeddelandena kan skiljer sig åt mellan olika parter och kan innehålla olika tillagd information eller till och med vara i olika format. En model tränas genom att använda kontrast-inlärning på genererade par av affärer och konfirmationer av affärer för att kunna para ihop paren i det latenta rummet genom att se vilka grannar som ligger närmast. Nogrannheten mäts genom att dela antalet korrekt klassificerade exempel med det totala antalet par i en grupp test-par. Modellen uppnår en noggrannhet så hög som 97.8% på 100 affärs-konfirmationspar med ett 30-dimensionellt latent rum, och det visas att kontrast-inlärning kan användas för att lösa problemet. Det är dock svårt att säga mycket om hur väl modellen kan generalisera de inlärda kunskaperna eftersom träningsdatan behövde genereras och därför saknar en del av komplexiteten av ett äkte data set.

So Ting Pong. / Summary in Chinese. / Thesis (M.B.A.)--Chinese University of Hong Kong. / Bibliography: leaves 150-151.

This study was conducted to explore budget DIY VR-gloves with haptics, and its use cases within Virtual Reality (VR). The VR-gloves were built according to an open-source project, called LucidGloves. A user test with 19 users was later conducted, where the VR-gloves were compared to Vive-controllers regarding the user experience in a VR-environment. The result for the entire test group showed no significant difference between the two units. A significant difference was found for the males, where the virtual hands were perceived more as their own (virtual body ownership) when using the VR-gloves. For the experienced VR-users, there was a significant difference regarding the object interaction, where the Vive-controllers appeared more realistic. The conclusions made is that if VR-gloves increase virtual body ownership for males, there is a potential to increase performance in VR-training simulators with VR-gloves. It is also crucial that the VR-gloves perform faultless to keep the user’s presence in VR. This is at the same time is difficult to do with one pair of gloves that should suit multiply hand sizes. The suggestions for improvement that were identified was to develop the responsivity and interaction for the VR-gloves. Further areas of improvement were to increase the robustness of the gloves, where suggestions were developed regarding a different component layout, battery placement and cord attachment. / Denna studie utfördes för att utforska egenbyggda billiga VR-handskar med haptik och dess användningsmöjligheter inom Virtual Reality (VR). VR-handskarna byggdes utifrån ett open source-projekt kallat LucidGloves. Ett användartest med 19 deltagare jämförde sedan VR-handskarna med Vive-kontroller i avseende på användarupplevelse i en VR-miljö. För hela testgruppen fanns ingen signifikant skillnad mellan de två styrsystemen. För männen upplevdes de virtuella händerna mer som ens egna händer (virtuellt kroppsägande) när VR-handskar användes. För enbart erfarna VR-användare fanns dessutom en signifikant skillnad gällande objektinteraktion, där det upplevdes mer realistiskt med Vive-kontrollerna. Slutsatser som dragits är att handskarna med ökat virtuellt kroppsägande för män visar potential i att förbättra träningssimulatorer i VR. Det är även viktigt att handskarna fungerar felfritt för att behålla användarens närvaro i VR (presence), men att detta är svårt att åstadkomma med ett par handskar som ska passa för olika handstorlekar. Förbättringsförslag som identifierades var framför allt att utveckla responsiviteten och interaktionen för handskarna. Ytterligare förbättringsförslag var att förbättra robustheten på handskarna, där förslag om förbättrad komponentlayout, batteriplacering och sladdfäste togs fram.

Laminar flow control holds major promise for increasing aircraft efficiency and increasing laminar flow over aerodynamic surfaces could decrease drag by up to 30 percent. The Flight Research Lab at Texas A&M University has studied laminar flow over a wing with 30 degrees of leading edge sweep with Discrete Roughness Elements (DREs) installed and has indicated that DREs can be used to increase laminar flow at Reynolds numbers up to 7.5 million at Mach 0.3. A new project, termed SARGE, has been commissioned in conjunction with NASA for studying DREs on a swept wing glove at conditions relevant to jet transports. The SARGE project must have an instrumentation system capable of accurately measuring flow conditions and transition location on the suction side of the glove. Infrared (IR) thermography has been selected as the primary transition detection tool. A heat transfer analysis has shown that solar radiation will warm the surface of the glove above the adiabatic wall temperature and therefore the laminar region will appear to be warmer. The FLIR SC8000 IR camera has been selected for this application due to its ability to produce high-resolution images in the appropriate IR band. High quality air data is also required for the experiment. A five-hole probe will be used to measure flow angle and velocity near the glove. This instrument will provide meanflow conditions due to its limited frequency response. High quality pressure transducers coupled with careful probe calibration will allow for differential measurements to be made with an uncertainty of +/- 0.03 degrees. Static pressure ports and high frequency response Kulite transducers will also be employed. Hotfilm sensors will be used to verify the state of the boundary layer on the glove through spectral analysis. A unique hotfilm array has been proposed that will enable the measurement of traveling wave vectors through a spectral technique. An experiment on the Flight Research Lab's Cessna O-2 to investigate the veracity of this technique has also been suggested. Thermocouples will also be installed on the glove's surface to monitor temperatures and verify transition location. The layout of the hotfilms and thermocouples is also detailed.

Demonstration of spanwise-periodic discrete roughness element laminar flow control (DRE LFC) technology at operationally relevant flight regimes requires extremely stable flow conditions in flight. A balance must be struck between the capabilities of the host aircraft and the scientific apparatus. A safe, effective, and efficient flight experiment is described to meet the test objectives, a flight test technique is designed to gather research-quality data, flight characteristics are analyzed for data compatibility, and an experiment is designed for data collection and analysis. The objective is to demonstrate DRE effects in a flight environment relevant to transport-category aircraft: [0.67 – 0.75] Mach number and [17.0M – 27.5M] Reynolds number. Within this envelope, flight conditions are determined which meet evaluation criteria for minimum lift coefficient and crossflow transition location. The angle of attack data band is determined, and the natural laminar flow characteristics are evaluated. Finally, DRE LFC technology is demonstrated in the angle of attack data band at the specified flight conditions. Within the angle of attack data band, a test angle of attack must be maintained with a tolerance of ± 0.1° for 15 seconds. A flight test technique is developed that precisely controls angle of attack. Lateral-directional stability characteristics of the host aircraft are exploited to manipulate the position of flight controls near the wing glove. Directional control inputs are applied in conjunction with lateral control inputs to achieve the desired flow conditions. The data are statistically analyzed in a split-plot factorial that produces a system response model in six variables: angle of attack, Mach number, Reynolds number, DRE height, DRE spacing, and the surface roughness of the leading edge. Predictions on aircraft performance are modeled to enable planning tools for efficient flight research while still producing statistically rigorous flight data. The Gulfstream IIB aircraft is determined to be suitable for a laminar flow control wing glove experiment using a low-bank-angle-turn flight test technique to enable precise, repeatable data collection at stabilized flight conditions. Analytical angle of attack models and an experimental design were generated to ensure efficient and effective flight research.

Research has shown that maximal exercise has a significant effect on cells of the immune system. Specifically, lymphocyte count increases during exercise and decreases to a value lower than baseline following an acute exhaustive bout of exercise. The overall lymphocyte response is well characterized, however, the ability of exercise to affect lymphocyte subfractions is unknown to our knowledge. The purpose of this study was to assess and evaluate the affects of a simulated tennis match across two sessions on lymphocyte subsets. Initial measurements such as age, height, weight, skinfold analysis, and heart rate were recorded for each player, as well as blood samples being obtained by a finger prick before and after the tennis sessions. The tennis protocol started with five serves to the deuce court and five serves to the ad court, then individuals hit twenty-four forehands and twenty-four backhands against an oscillating ball machine. Each bout of serves and ground strokes were repeated ten times, with one minute rests in between each session. Before and immediately after completing the tennis trial, subjects were pricked with a lancet on the non dominant hand so to obtain at least two capillary tubes of blood. Whole blood was then added to the antibody cocktail, which is mixed according to the antibodies that were tested, which were CD4, CD8, CD19, CD95, and CX3CR1. Whole blood was added to red blood cell lysis buffer and fixation buffer, and the blood solution was incubated with antibodies specific to cell phenotype. The main results of this study indicated that there was a decrease in mainly post cell counts in pre and post CD19/CD95 measurements (P= .007), an increase in CD8/CX3CR1 in pre counts and an increase then decrease in post counts without wearing the bionic glove (P= .042), and a decrease in CD4 in the post count measurement with the bionic tennis glove (P= .043). The study’s can assist in making recommendations for after match treatment such as health and diet suggestions. Knowledge of prevention and treatment methods are low in the field of tennis and immune functions, so findings in this area could prevent elite athletes from contracting infections between matches.

Not available / text

The repetitive and sometimes mundane nature of conventional rehabilitation therapy provides an ideal opportunity for development of interactive and challenging therapeutic games that have the potential to engage and motivate the players. Different game design techniques can be used to design rehabilitation games that work alongside robotics to provide an augmentative therapy to stroke patients in order to increase their compliance and motivation towards therapy. The strategy we followed to develop such a system was to (i) identify the key design parameters that can influence compliance, prolonged activity, active participation and patient motivation, (ii) use these parameters to design rehabilitation games for robot-mediated stroke-rehabilitation, (iii) investigate the effects of these parameters on motivation and performance of patients undergoing home-based rehabilitation therapy. Three main studies were conducted with healthy subjects and stroke subjects. The first study identified the effects of the design parameters on healthy players' motivation. Using the results from this study, we incorporated the parameters into rehabilitation games, following player-centric iterative design process, which were formatively evaluated during the second study with healthy subjects, stroke patients, and health-care professionals. The final study investigated the research outcomes from use of these games in three patient's homes during a 6 weeks clinical evaluation. In summary, the research undertaken during this PhD successfully identified the design techniques influencing patient motivation and adherence as well as highlighted further important elements that contribute to maintaining therapeutic interaction between patients and the therapy medium, mainly the technological usability and reliability of the system.

Exposure to dust and latex allergens has been well established as a cause of health problems but there have been few studies that have examined the extent of problems in latex glove manufactures, especially in Thailand. Therefore, the overall aim of the study was to determine whether exposure to dust and latex allergens was associated with health problems in Thai latex glove manufacturing workers.The study was conducted in 3 latex glove factories in southern Thailand that employed 1,481 workers (931 Thai and 550 foreign workers). Full shift personal air sampling of workers in 10 different departments was carried out using the IOM Multidust™ sampling Head equipped with 25 mm 1.0 mm Polytetrafluoroethyene filters at a flow rate of 2 L/min. A commercial ELISA kit was employed to quantify levels of the Hev b 6.02 latex allergen in air sample extracts. Information on demographics and, occupational history, exposures and health related problems attributed to work was collected by questionnaires completed by Thai workers. Lung function testing using a spirometer was also conducted among Thai workers. Univariate linear mixed-model analysis was used to determine differences in the dust and specific latex allergen concentrations between factories, departments, shifts and days. Multivariate linear mixed model analysis was used to determine exposure determinants to estimate exposure concentrations of inhalable dust and specific latex allergen. Logistic regression analysis was used to determine association between exposure to inhalable dust and specific latex allergens.292 air samples were collected. The geometric mean (GM) and a range of inhalable dust was 0.88 (0.01 – 12.34) mg/m3. The highest dust levels were found in the Stripping 3.01, (0.73 – 12.34) mg/m3 and lowest in the Office 0.11, (0.06 – 0.92 mg/m3) departments. Hev b 6.02 levels were 9.35 (4.08 – 345.53) ng/m3 and the highest and lowest levels were also found in the Stripping [19.76, (4.72 – 192.90) ng/m3] and Office [5.60, (4.37 – 14.64) ng/m3] departments. Factory and Department were important determinants of both inhalable dust and Hev b 6.02levels. 660 (157 men, 503 women) out of 931 workers completed a self-reported questionnaire. The prevalence of health problems attributed to work was highest for skin problems (25.0%) followed by Rhinitis (23.9%), conjunctivitis (22.5%) and cough (16.2%). Cumulative exposure to dust was associated with conjunctivitis attributed to work (OR and 95%CI = 1.02, [1.00 - 1.03]) after adjusted for confounders. A negative association was found between rhinitis and exposure to average dust level (OR and 95%CI =0.69 [0.48 - 0.99]). No other associations were found between either exposure to dust and self-reported ill-health. There were also no association between exposure to Hev b 6.02 and ill-health although skin problems were associated with the highest Hev b 6.02 level (OR and 95% CI = 1.72 [1.02 - 2.91]). Lung function testing was conducted in 474 Thai workers; 41 had airway restriction, 3 airway obstruction and 1 with a combined problem. Cumulative exposure to dust was associated with airway restriction after adjustment for confounders (OR and 95% CI = 1.02 (1.00 - 1.04).

This thesis investigates how to remove personal data from logs using machine learning when rule-based scripts are not enough and manual scanning is too extensive. Three types of machine learning models were created and compared. One word model using logistic regression, another word model using LSTM and a sentence model also using LSTM. Data logs were cleaned and annotated using rule-based scripts, datasets from various countries and dictionaries from various languages. The created dataset for the sentence based model was imbalanced, and a lite version of data augmentation was applied. A hyperparameter optimization library was used to find the best hyperparameter combination. The models learned the training and the validation set well but did perform worse on the test set consisting of log data from a different server logging other types of data. / Detta examensarbete undersöker om det är möjligt att skapa ett program som automatiskt identifierar och tar bort persondata från dataloggar med hjälp av maskinlärning. Att förstå innebörden av vissa ord kräver också kontext: Banan kan syfta på en banan som man kan äta eller en bana som man kan springa på. Kan en maskinlärningsmodell ta nytta av föregående och efterkommande ord i en sekvens av ord för att få en bättre noggrannhet på om ordet är känsligt eller ej. Typen av data som förekommer i loggarna kan vara bland annat namn, personnummer, användarnamn och epostadress. För att modellen ska kunna lära sig att känna igen datan krävs det att det finns data som är färdigannoterad med facit i hand. Telefonnummer, personnummer och epostadress kan bara se ut på ett visst sätt och behöver nödvändigtvis ingen maskininlärning för att kunna pekas ut. Kan man skapa en generell modell som fungerar på flera typer av dataloggar utan att använda regelbaserade algoritmer. Resultaten visar att den annoterade datan som användes för träning kan ha skiljt allt för mycket från de loggar som har testats på (osedd data), vilket betyder att modellen inte är bra på att generalisera.