教学课程资源库 更多>>

关键词： Biomedical engineering   Computer science   Artificial intelligence   Neurosciences  

摘要： Parkinson’s disease (PD) is a neurodegenerative disease that affects motor abilities with increasing severity as the disease progresses. Traditional methods for diagnosing PD require specialists scoring qualitative symptoms using the motor subscale of the Unified Parkinson’s Disease Rating Scale (UPDRS-III). Using force-plate data during quiet standing (QS), this study uses machine learning to target the characterization and prediction of PD and UPDRS-III. The purpose of predicting different subscores of the UPDRS-III is to give specialists more tools to help make an informed diagnosis and prognosis. The classification models employed classified PD with a sensitivity of 87.5% and specificity of 83.1%. Stepwise forward regression indicated that features correlated with base of support were most useful in the prediction of head rigidity (r-square = .753). Although there is limited data, this thesis can be used as an exploratory study that evaluates the predictability of UPDRS-III subscores using QS data. Similar prediction models can be implemented to a home setting using low-cost force plates as a novel telemedicine technique to track disease progression.

关键词： Physiology   Public health   Biostatistics   Computer science  

摘要： COVID-19 is an infectious disease caused by a novel human coronavirus, SARS-CoV-2. Patients with pre-existing COVID-19 infection are vulnerable and more proned to other health disorders like respiratory failure, multiple organ dysfunction etc. Acute kidney injury (AKI) is also strongly associated with hospitalized COVID-19 patients, having poor health outcomes. Hence it would be beneficial to predict AKI and analyze the key predictors among the COVID-19 patients. Although several research studies have been conducted in predicting the diseases based on different machine learning algorithms, but very few have been reported to predict the AKI and recovery of the AKI patients. Hence to address the problem of predicting the above-mentioned outcomes, an XGBoost machine learning model was developed for each prediction task that takes admission values recorded within 48 hours of the patient’s admission. A dataset of 3043 hospitalized patients was considered which was collected at Stony Brook University Hospital during the current COVID-19 ***, studies predicting whether the patient with and without COVID-19 undergoes renal replacement therapy (RRT) or dies in the hospital were also conducted. Our machine learning models were able to predict COVID-19-associated AKI, RRT, death with an average precision (AP) of 0.36, 0.54 and 0.66 respectively while their baselines were 0.18, 0.03 and0.10. In those without COVID-19 the APs were 0.23, 0.24, 0.35 while their baselines were 0.09, 0.004 and 0.05. To have enhanced APs, the hyper-parameters of the models were tuned and their interpretability is reported using SHAP. Among COVID-19 patients, the key predictors for AKI include serum creatinine (Scr), age, urine RBCs, for RRT they were Ionized Ca, IL-6, Na urine, Scr and for death they were arterial O2, fraction of inspired oxygen, lymphocyte count. Similar key predictors can be defined for COVID-19 negative patients.

关键词： Energy   Atmospheric sciences   Computer science  

关键词： Public health   Gerontology   Computer science   Aging  

摘要： Covid-19 and the increasing population of older adults in developed countries have increased the need for alternative solutions for preventative health and long-term care. Smart homes and other assistive technology have been proposed as a solution to detect health problems early and prevent major issues later, provide better care access through telemedicine solutions, and support the autonomy of older adults with cognitive or physical issues who wish to continue to live in their homes. Rural older adults’ acceptance of assistive technology is understudied compared to other groups, but the few studies available indicated that many tend to reject it. However, Covid-19 may increase interest in the technology, and the goal of this project was to determine if rural Oklahoman older adults’ acceptance of the technology has increased enough for it to be implemented. This is a qualitative study that used Grounded Theory coding to determine common themes of participants to determine the level of acceptance. Interviews indicated that acceptance is still low for some devices but that rural older Oklahomans may consider other remote health instruments for specific conditions, fall detection, and home security devices if recommended by a doctor and if their issues with the technology, such as self-image, independence, and financial concerns, were resolved. If the acceptance of these devices is increased in the future, it may be possible to provide better preventative care for rural Oklahomans, detecting problems early and treating the main issues to delay major health complications. This could narrow healthcare disparities between urban and rural older adults, especially during special situations like pandemics.

关键词： computer   teaching   learning   university   Web of Science   student  

摘要： Among the tools used in teaching and learning, computers are very important. Determining what kind of computer-related studies are carried out in the learning and teaching processes for university students will contribute to future research. In this context, the aim of the study is to examine university students' documents related to computers in teaching and learning, scanned in the Web of Science database, thematically and methodologically. The study was designed by adopting the case study method from the qualitative research models. A sample was not determined for the articles to be included in the study but it was aimed to reach the whole population. In this context, all documents in the Web of Science database have been accessed through the library system of the university. In the document search, the keywords 'learning and teaching' and 'computer' were searched in all areas, and the studies were found by searching the keyword 'university student' within the title. All the studies were analysed by year, document type, Web of Science category, country and language of publication. The findings obtained from the research are explained in detail by content analysis method. A total of 305 documents were reached in the research findings. The findings obtained as a result of the research carried out are given in detail in the findings and results section.

关键词： Engineering   Computer science  

摘要： Code summarization and generation are valuable tasks to master for their wide range ofapplications in code readability and code translation to name a few. This research work is an extension of previously conducted research on the use of PLBART, a sequence-tosequence transformer model used for a variety of program and language understanding and generation (PLUG) tasks. The ultimate goal is to improve the performance of PLBART by modifying the noise function of it's denoising autoencoder. The current noise function corrupts code tokens randomly, but we hope to improve performance by masking nodes on the corresponding Syntax Tree (AST) instead. To integrate the AST structure into the self-attention mechanism, we adopt the dependency-guided self-attention mechanism explored in NLP literature in particular [ZKC21]. However, from the AST, we cannot compute distances between all tokens that appear in a code since they need not necessarily appear in the parse tree structure. So, we investigate how we can derive distances between tokens from the AST structure.

关键词： Mechanical engineering   Mechanics   Computer science  

摘要： Two approaches for computational contact mechanics are explored to provide an accurate and robust numerical simulation of frictional contact. First, a novel quasi-static contact algorithm (Method of First Violation) is developed, which provides a solution over the load history that is admissible in the sense that all equality and inequality constraints are satisfied at each load step and every point between load steps. The central idea of this algorithm is to calculate each load increment to be exactly which is necessary to cause the first nodal inequality constraint to be violated. At the end ofthat load increment, the status of the associated node is switched to a different status (such as from contact to non-contact) where the inequality constraint is replaced with an equality constraint, and a new inequality constraint is introduced. Importantly, the constraints of the previous and the new node status are simultaneously satisfiedat that point in the loading cycle. The Method of First Violation (MFV) algorithm is illustrated using a compliance formulation for degrees of freedom on the surface of a half-plane loaded by a rigid indenter with friction. The compliance formulation permits considering only the degrees of freedom in the vicinity of contact, therefore extremely fine surface discretization is possible using only a modest number of degrees of freedom. Because this contact problem has been well studied using analytic techniques and because of the extremely fine mesh resolution, it is possible to perform a critical comparison with analytical solutions. An added benefit of this algorithm is its greatly improved robustness. This indenter/half-plane discretization was also used to test MFV where friction was represented by a two-parameter (static and kinematic) model. Though such problems are notoriously ill-conditioned, the MFV remained robust. It was also used to explore a problem involving the Dahl friction model. Second, a compliance matrix formulation is d

关键词： Computer science   Artificial intelligence  

摘要： The focus of this research for my thesis is filtering techniques as a precursor for the analysis of DeepFakes. In this research, the exploitation analysis of OpenCV local descriptor filters and frequency, or also known as wavelet, decomposition are analyzed. Local descriptors and wavelets both exploit different aspects in images. The local descriptors work as indicators of unique features in an image; meanwhile, wavelets emphasize underlying features within said image. Multiple filtering techniques were examined for this research topic to find suitable techniques for this tampered media, DeepFakes. This included OpenCV image matching techniques (local descriptors) and frequency (wavelet) techniques. This research does not exclude intermediate research using single techniques from both local descriptors and wavelets techniques for primary comparison. Scale-Invariant Feature Transform and Discrete Wavelet Transform were used for testing of the beneficial nature of these methods. In previous research, the focus was on the standard deviation of the points of interest from local descriptors like Scale-Invariant Feature Transform to detect whether visual media has been compromised with misinformation. Frequency is often used to look into underlying features within an image. This technique of using this form of decomposition can expand to media other than images, but possible face recognition for the identity aspect of the DeepFakes. DeepFakes are becoming a trend, and hence there is a bigger determination to be able to detect this fake media. Two types of filtering techniques are analyzed to examine the usability of layering filtering like local descriptors or/and wavelet-based decomposition as a beneficial technique to bring out more information from DeepFake data to further detect if tampered.

关键词： Computer science   Computer engineering   Remote sensing   Artificial intelligence  

摘要： The age of automation has led to significant research in the field of Machine Learning and Computer Vision. Computer Vision tasks fundamentally rely on information from digital images, videos, texts and sensors to build intelligent systems. In recent times, deep neural networks combined with computer vision algorithms have been successful in developing 2D object detection methods with a potential to be applied in real-time systems. However, performing fast and accurate 3D object detection is still a challenging problem. The automotive industry is shifting gears towards building electric vehicles, connected cars, sustainable vehicles and is expected to have a high growth potential in the coming years. 3D object detection is a critical task for autonomous driving vehicles and robots as it helps moving objects in the scene to effectively plan their motion around other objects. 3D object detection tasks leverage image data from camera and/or 3D point clouds obtained from expensive 3D LiDAR sensors to achieve high detection accuracy. The 3D LiDAR sensor provides accurate depth information that is required to estimate the third dimension of the objects in the scene. Typically, a 64 beam LiDAR sensor mounted on a self-driving car cost around $75000. In this thesis, we propose a cost-effective approach for 3D object detection using a low-cost 2D LiDAR sensor. We collectively use the single beam point cloud data from 2D LiDAR for depth correction in pseudo-LiDAR. The proposed methods are tested on the KITTI 3D object detection dataset.

关键词： Computer science   Artificial intelligence  

摘要： Smart Home IoT is gaining popularity because of its ability to render a connected experience and a high level of automation to its users. To render this connected experience, smart home devices need to collect and share data from their environments. From privacy and security standpoint, the data collection can be an important cause of concern for smart home IoT users. The research presented in this dissertation is focused on understanding how smart home IoT users make privacy decisions. With the understanding of these decision preferences, privacy settings interfaces for smart home users are created which can be helpful in setting privacy preferences effectively. In this dissertation, privacy decision making in smart home IoT is investigated from three angles: First, understanding how contextual factors such as entities collecting/receiving data and storage of location influences privacy decisions. Second, investigating how factors like heuristics (in form of defaults and framing) and personality characteristics which lie outside of decision making context influence privacy decisions. Third, how do conceptual models associated with smart homes influence the privacy management experience of smart home IoT users. In a controlled experiment which presented participants with multiple contextual scenarios, the data analysis showed that the participants tend to emphasize some contextual factors more over the others and that their decision making is influenced by heuristics like defaults and framing. The regression modeling results of privacy decisions informed the design of privacy settings interfaces which can be used to manage privacy decision in smart homes. By using machine learning methods, participants were clustered on the basis of similarity in their privacy decisions. Upon further analysis of these clusters, it was observed that the interface design needs of participants varied across different clusters. This observation led to the creation of personalized privac