Geologic sequestration of anthropogenic carbon dioxide (CO2) greenhouse gas emissions is an engineering solution that potentially reduces CO2 emissions released into the atmosphere thereby limiting their effect on climate change. This study focuses on Woodside field as a potential storage and sequestration site for CO2 emissions. The Woodside field is positioned on a doubly plunging, asymmetrical anticline on the northeast flank of the San Rafael Swell. Particular focus will be placed on the Permian White Rim Sandstone, Black Box Dolomite and Black Dragon Member of the Triassic Moenkopi Formation as the reservoir/seal system to store and sequester CO2 at Woodside field. The White Rim Sandstone, the primary target reservoir, is divided into three stratigraphic intervals based on facies analysis: a lower sand sheet facies (about 60 ft or 18 m), a thick middle eolian sandstone facies (about 390 ft or 119 m), and an upper marine reworked facies (about 70 ft or 21 m). Porosity and permeability analyses from the outcrop indicate good reservoir quality in the eolian sandstone and reworked facies. Porosity in the White Rim Sandstone ranges from 7.6 to 24.1% and permeability reaches up to 2.1 D. The maximum combined thickness of the three facies is 525 ft (160 m) at Woodside field providing a significant volume of porous and permeable rock in which to store CO2. The Black Box Dolomite is the secondary potential reservoir for CO2 storage at Woodside field and has a gross thickness up to 76 ft (23 m). The Black Box Dolomite is divided into four lithofacies: a basal nodular dolomudstone (8.2 -15 ft or 3.5-4.5 m), a dolowackestone (25-37 ft or 7.5-11 m), a dolomitic sandstone (0-8.2 ft or 0-2.5 m), and an upper sandy dolowackestone (0-16 ft or 0-4.9 m). Porosity and permeability analyses indicate reservoir potential in the dolowackestone, dolomitic sandstone, and sandy dolowackestone lithofacies. Porosity in the Black Box Dolomite ranges from 6.6 to 29.2% and permeability reaches up to 358 mD. The nodular dolomudstone lithofacies has relatively poor reservoir quality with porosity up to 9.4% and permeability up to 0.182 mD. This lithofacies could act as a baffle or barrier to fluid communication between the White Rim Sandstone and Black Box Dolomite. The Black Dragon Member of the Triassic Moenkopi Formation will serve as the seal rock for the relatively buoyant CO2 stored in the underlying formations. The Black Dragon Member is comprised of four lithofacies: a chert pebble conglomerate; an interbedded sandstone, siltstone, and shale; a trough cross-stratified sandstone, and an oolitic and algal limestone. The Black Dragon Member has a maximum thickness of 280 ft (85 m) at Woodside field. Mudstone beds contain from 0.16 to 0.47% porosity. QEMSCAN analysis indicates several minerals within shale beds that may react with a CO2-rich brine including calcite (18.73 to 23.43%), dolomite (7.56 to 7.89%), alkali feldspar (4.12 to 4.43 %), glauconite (0.04 to 0.05%), and plagioclase (0.03 to 0.04%). Silty mudstones comprise 75% of this member at Black Dragon Canyon. Volumetric estimates for Woodside field were calculated based on the 10th, 50th, and 90th percent probabilities (P10, P50, and P90). The White Rim Sandstone is the primary target reservoir and has capacity to hold 2.2, 8.8, or 23.7 million metric tonnes (P10, P50, and P90 respectively) of CO2 within the structural closure of Woodside field. The Black Box Dolomite may hold 0.5, 1.8, or 4.5 million metric tonnes, respectively of additional CO2 within the structural closure of Woodside field. These two formations combined have the capacity to store up to 28.3 million metric tonnes (P90) of CO2.

To determine whether a credit limit for a corporate client should be changed, a financial institution writes a PM containingtext and financial data that then is assessed by a credit committee which decides whether to increase the limit or not. To make thisprocess more efficient, machine learning algorithms was used to classify the credit PMs instead of a committee. Since most machinelearning algorithms are black boxes, the LIME framework was used to find the most important features driving the classification. Theresults of this study show that credit memos can be classified with high accuracy and that LIME can be used to indicate which parts ofthe memo had the biggest impact. This implicates that the credit process could be improved by utilizing machine learning, whilemaintaining transparency. However, machine learning may disrupt learning processes within the organization. / För att bedöma om en kreditlimit för ett företag ska förändras eller inte skriver ett finansiellt institut ett PM innehållande text och finansiella data. Detta PM granskas sedan av en kreditkommitté som beslutar om limiten ska förändras eller inte. För att effektivisera denna process användes i denna rapport maskininlärning istället för en kreditkommitté för att besluta om limiten ska förändras. Eftersom de flesta maskininlärningsalgoritmer är svarta lådor så användes LIME-ramverket för att hitta de viktigaste drivarna bakom klassificeringen. Denna studies resultat visar att kredit-PM kan klassificeras med hög noggrannhet och att LIME kan visa vilken del av ett PM som hade störst påverkan vid klassificeringen. Implikationerna av detta är att kreditprocessen kan förbättras av maskininlärning, utan att förlora transparens. Maskininlärning kan emellertid störa lärandeprocesser i organisationen, varför införandet av dessa algoritmer bör vägas mot hur betydelsefullt det är att bevara och utveckla kunskap inom organisationen.

AI is entering clinical care and the healthcare sector in a big way, at the same time, a growing number of scholars are concerned that this technology cannot adhere to current bioethical principles. In particular, there are increasing concerns that AI poses a threat to the autonomy of patients by being irreconcilable with the practice of informed consent. In this essay, I shall defend the thesis that some applications of AI can be reconciled with a revised version of informed consent – what I call AI Adapted Informed Consent. This solution shall not rest on the idea of making black box AI more transparent or explicable. Instead, I shall argue that black box AI does not necessarily withhold the kind of information necessary for informed consent. Rather, patients can be given epistemic access to the kind of information necessary to make an informed decision, as well as being informed as to how the AI is used in the medical decision-making and in the assessment of their medical situation. Hence, this solution offers a re-interpretation of informed consent as information about contextual functioning and role of AI in medical decision-making. Drawing on republican interpretations of freedom as nondomination, I argue that demands for informed consent can only be restrained if it preserves the voluntariness of our decisions. Hence, I shall conclude that my adapted informed consent thesis allows for the possibility that some applications of black box AI in clinical care can be reconciled with informed consent and due respect for patient autonomy – if three specific conditions can be met.

Several exact and approximate dynamic programming formulations have already been proposed to solve hotel revenue management (RM) problems. To obtain tractable solutions, these methods are often bound by simplifying assumptions which prevent their application on large and dynamic complex systems. This dissertation introduces HotelSimu, a flexible simulation-based optimization approach for hotel RM, and investigates possible approaches to increase the efficiency of black-box optimization methods in the presence of noise. In fact, HotelSimu employs black-box optimization and stochastic simulation to find the dynamic pricing policy which is expected to maximize the revenue of a given hotel in a certain period of time. However, the simulation output is noisy and different solutions should be compared in a statistically significant manner. Various black-box heuristics based on variations of random local search are investigated and integrated with statistical analysis techniques in order to manage efficiently the optimization budget.

No description available.

No description available.

Cieľom tejto práce je vývoj nástroja na automatické generovanie testov pre aplikácie s grafickým užívateľským rozhraním v~prostredí GNOME. Na generovanie testov sú použité metadáta asistenčných technológií, konrétne AT-SPI. Navrhnutý generátor testov využíva dané metadáta na vytvorenie modelu testovanej aplikácie. Model mapuje sekvencie udalostí, ktoré generátor vykoná na testovanej aplikácii počas generovania testov. Súčasťou procesu generovania je zároveň detekcia závažných chýb v testovaných aplikáciách. Výstupom procesu generovania sú automatizované testy, ktoré sú vhodné na regresné testovanie. Funkčnosť implementovaného generátora testov bola úspešne overená testovaním 5 aplikácií s otvoreným zdrojovým kódom. Počas testovania aplikácií navrhnutým nástrojom sa preukázala schopnosť detekovať nové chyby.

Zariadenia, na ktoré sa každodenne spoliehame, sú stále zložitejšie a využívajú zložitejšie protokoly. Jedným z týchto protokolov je Wi-Fi. S rastúcou zložitosťou sa zvyšuje aj potenciál pre implementačné chyby. Táto práca skúma Wi-Fi protokol a použitie fuzz testingu pre generovanie semi-validných vstupov, ktoré by mohli odhaliť zraniteľné miesta v zariadeniach. Špeciálna pozornosť bola venovaná testovaniu Wi-Fi v systéme ESP32 a ESP32-S2. Výsledkom práce je fuzzer vhodný pre testovanie akéhokoľvek Wi-Fi zariadenia, monitorovací nástroj špeciálne pre ESP32 a sada testovacích programov pre ESP32. Nástroj neodhalil žiadne potenciálne zraniteľnosti.

Model-based testing (MBT) is a black-box software testing technique that focuses on specification of the system-under-test (SUT) and/or its environment. It uses models to automatically generate a large number of tests. To the best of our knowledge, no study has investigated the correlation of model coverage with SUT coverage using more advanced coverage criteria (such as edge-pair coverage) and the correlation of coverage (at model level and SUT level) with test suite effectiveness using non-adequate test suites in the context of MBT despite the prominence of non-adequate test suites in industry. To carry out the investigation, we extend an existing open-source MBT tool called Modbat to measure edge-pair coverage at model level, implement a new tool called PaCovForJbc to measure edge-pair coverage, edge coverage and node coverage at SUT level. Finally, we perform an experiment using these tools applied on three projects: “ArrayList”, and “LinkedList” of Java standard library, and “Apache ZooKeeper”. Overall, the results suggest the following: Edge and edge-pair coverage at model level often have a moderate to high correlation with the same type of coverage at SUT level, while that link between model and SUT for node coverage is weaker. Moreover, coverage criteria at SUT level often have a moderate to high correlation with test suite effectiveness, and a coverage criterion at SUT level has a slightly higher correlation with test suite effectiveness than the same type of coverage at model level. Regarding coverage at model level, edge and edge-pair coverage at model level have a slightly higher correlation with test suite effectiveness than node coverage at model level. Note that the mentioned suggestions need to be taken with discretion, because results vary depending on the project and/or coverage criterion under investigation. / Modellbaserad testning (MBT) är en black-box-testteknik som fokuserar på specifikation av system-under-test (SUT) och/eller dess miljö. MBT använder modeller för att generera ett stort antal tester automatiskt. Såvitt vi vet, finns ingen studie som undersökt korrelationen mellan modelltäckning och täckning av SUT med hjälp av mer avancerade täckningskriterier såsom kantparstäckning. Dessutom finns ingen studie som undersökt korrelationen mellan täckning (på modellnivå och SUT-nivå) och effektivitet av icke- adekvata testsviter som genereras med hjälp av MBT trots betydelsen av icke-adekvata testsviter i industrin. För att utföra undersökningen, utökar vi ett ”open-source” MBT-verktyg som kallas för Modbat för att mäta kantparstäckning på modellnivå. Dessutom implementerar vi ett nytt verktyg som kallas för PaCovForJbc för att mäta kantpars-, kant- och nodtäckning på SUT-nivå. Till slut utför vi experiment genom att applicera Modbat och PaCovForJbc på tre projekt: ”ArrayList” och ”LinkedList” av Javas standardbibliotek samt ”Apache ZooKeeper”. Sammantaget indikerar resultaten följande: Kant- och kantparstäckning på modellnivå har ofta en måttlig till hög korrelation med samma typ av täckning på SUT- nivå, medan länken mellan modell och SUT för nodtäckning är svagare. Dessutom har täckningskriterier på SUT-nivå ofta en måttlig till hög korrelation med testsvitseffektivitet, och ett täckningskriterium på SUT-nivå har en aning högre korrelation med testsvitseffektivitet än samma typ av täckning på modellnivå. Angående täckning på modellnivå har kant- och kantparstäckning på modellnivå en aning högre korrelation med testsvitseffektivitet än nodtäckning på modellnivå. Observera att de nämnda förslagen måste tas med diskretion, eftersom resultaten varierar beroende på projektet och/eller täckningskriteriet som undersöks.

Insurance is built on the principle that a group of people contributes to a common pool of money which will be used to cover the costs for individuals who suffer from the insured event. In a competitive market, an insurance company will only be profitable if their pricing reflects the covered risks as good as possible. This thesis investigates the recently proposed Combined Actuarial Neural Network (CANN), a model nesting the traditional Generalised Linear Model (GLM) used in insurance pricing into a Neural Network (NN). The main idea of utilising NNs for insurance pricing is to model interactions between features that the GLM is unable to capture. The CANN model is analysed in a commercial insurance setting with respect to two research questions. The first research question, RQ 1, seeks to answer if the CANN model can outperform the underlying GLM with respect to error metrics and actuarial model evaluation tools. The second research question, RQ 2, seeks to identify existing interpretability methods that can be applied to the CANN model and also showcase how they can be applied. The results for RQ 1 show that CANN models are able to consistently outperform the GLM with respect to chosen model evaluation tools. A literature search is conducted to answer RQ 2, identifying interpretability methods that either are applicable or are possibly applicable to the CANN model. One interpretability method is also proposed in this thesis specifically for the CANN model, using model-fitted averages on two-dimensional segments of the data. Three interpretability methods from the literature search and the one proposed in this thesis are demonstrated, illustrating how these may be applied. / Försäkringar bygger på principen att en grupp människor bidrar till en gemensam summa pengar som används för att täcka kostnader för individer som råkar ut för den försäkrade händelsen. I en konkurrensutsatt marknad kommer försäkringsbolag endast vara lönsamma om deras prissättning är så bra som möjligt. Denna uppsats undersöker den nyligen föreslagna Combined Actuarial Neural Network (CANN) modellen som bygger in en Generalised Linear Model (GLM) i ett neuralt nätverk, i en praktiskt och kommersiell försäkringskontext med avseende på två forskningsfrågor. Huvudidén för en CANN modell är att fånga interaktioner mellan variabler, vilket en GLM inte automatiskt kan göra. Forskningsfråga 1 ämnar undersöka huruvida en CANN modell kan prestera bättre än en GLM med avseende på utvalda statistiska prestationsmått och modellutvärderingsverktyg som används av aktuarier. Forskningsfråga 2 ämnar identifiera några tolkningsverktyg som kan appliceras på CANN modellen samt demonstrera hur de kan användas. Resultaten för Forskningsfråga 1 visar att CANN modellen kan prestera bättre än en GLM. En literatursökning genomförs för att svara på Forskningsfråga 2, och ett antal tolkningsverktyg identifieras. Ett tolkningsverktyg föreslås också i denna uppsats specifikt för att tolka CANN modellen. Tre av tolkningsverktygen samt det utvecklade verktyget demonstreras för att visa hur de kan användas för att tolka CANN modellen.