教学课程资源库 更多>>

关键词： Computer science   Artificial intelligence  

摘要： A synthetic data set for kidney transplantation is developed using a Wasserstein generative adversarial network (WGAN) and a donor-recipient HLA matching algorithm. Like many medical data sets, much of the kidney transplant data set is categorical. A new method for dealing with categorical data in GANs is proposed and the results are analyzed. The real-valued data is prepared with a process similar to one-hot encoding, but instead of using ones and zeros, the values are logit in nature, with a large positive value corresponding to one and large negative value corresponding to zero. By capturing the logit distributions and correlations between categories within a GAN, the generator can create a synthetic version of the data that will resemble the real, un-encoded data set after a softmax function is applied and values are stochastically selected. While the statistical metric used demonstrated that the synthetic and real data sets did not come from the same distribution, a visual inspection shows general similarity between the two data sets. A rematching simulation performed on the real and synthetic data set shows relatively similar results.

关键词： Temporal logic   Systems design   Autonomous vehicles   Adaptation   Automobiles   Unmanned aerial vehicles   Design   Feedback   Software engineering   Robotics   Aerospace engineering   Computer science   Engineering   Logic   Transportation  

摘要： The design of autonomous systems is challenging and ensuring their trustworthiness can have different meanings, such as i) ensuring consistency and completeness of the requirements by a correct elicitation and formalization process; ii) ensuring that requirements are correctly mapped to system implementations so that any system behaviors never violate its requirements; iii) maximizing the reuse of available components and subsystems in order to cope with the design complexity; and iv) ensuring correct coordination of the system with its environment. Several techniques have been proposed over the years to cope with specific problems. However, a holistic design framework that, leveraging on existing tools and methodologies, practically helps the analysis and design of autonomous systems is still missing. This thesis explores the problem of building trustworthy autonomous systems from different angles. We have analyzed how current approaches of formal verification can provide assurances: 1) to the requirement corpora itself by formalizing requirements with assume/guarantee contracts to detect incompleteness and conflicts; 2) to the reward function used to then train the system so that the requirements do not get misinterpreted; 3) to the execution of the system by run-time monitoring and enforcing certain invariants; 4) to the coordination of the system with other external entities in a system of system scenario and 5) to system behaviors by automatically synthesize a policy which is correct.

关键词： Computer science   Electrical engineering   Sparsity   Accuracy   Deep learning   Teaching methods   Datasets   Scanners   Magnetic resonance imaging   Knee   Neural networks   Brain   Art   Energy   Performance evaluation  

摘要： Convolutional Neural Networks (CNNs) are highly effective for image reconstruction problems. Typically, CNNs are trained on large amounts of training images. Recently, however, un-trained CNNs such as the Deep Image Prior and Deep Decoder have achieved excellent performance for image reconstruction problems such as denoising and inpainting, without using any training data. In this work, we study the performance of un-trained neural networks for the reconstruction problem arising in accelerated Magnetic Resonance Imaging (MRI). For this purpose, we view the performance from two perspectives: reconstruction accuracy, and reconstruction reliability. Reconstruction accuracy: One of the main goals in solving image reconstruction tasks is obtaining a high-quality image. In order to measure the quality of the images reconstructed by an un-trained network, we first propose a highly-optimized un-trained recovery approach based on a variation of the Deep Decoder. Afterward, through extensive experiments, we show that the resulting method significantly outperforms conventional un-trained methods such as total-variation norm minimization, as well as naive applications of un-trained networks. Most importantly, the proposed approach achieves on-par performance with a standard trained baseline, the U-net, on the FastMRI dataset, a dataset for benchmarking deep-learning based reconstruction methods. This demonstrates that there is less benefit in learning a prior for solving the accelerated MRI reconstruction problem. This conclusion is drawn by comparison with a baseline trained neural network, however, state-of-the-art methods still slightly outperform U-net (and hence the proposed approach in this work). Reconstruction reliability: Recent works on accelerated MRI reconstruction suggest that trained neural networks are not reliable for image reconstruction tasks, albeit achieving excellent accuracy. For example, at the inference time, small changes (also referred to as adversaria

关键词： Engineering   Computer science   Robotics  

摘要： Localization of robotic systems is a necessity for control of robotic systems and often requires the use of costly sensors and equipment onboard the robot or installed within a facility. GPS or GNSS sensors are effective for localizing robots in GPS-accessible environments, but do not work as well in dense urban areas or inside fully enclosed buildings, such as factories or warehouses. Sensors such as LiDAR can be costly and require substantial experience and knowledge to utilize as well as potential changes to infrastructure for use. Computer vision-based localization systems offer potential as a localization solution for various applications. A low-cost overhead camera system was developed to localize robotic systems in an indoor facility, aiding the development and verification of algorithms. A visual servoing path following robot was built and developed utilizing Robot Operating System to examine the computer vision-based localization camera system. A track was designed and assembled, and low-cost cameras were mounted in an overhead configuration. A program to track the robots’ position was developed utilizing multiple camera feeds, open-source computer vision tools, and fiducial marker tracking. Utilizing the camera feed from three different cameras and ArUco fiducial markers, localization of a robotic system was conducted in static positions with an average of 0.80% error between the physical measurement and measurement made by the camera system. The path following robot was tracked and the RMS position acquired by the developed localization system compared to that captured by the robot’s onboard camera with an average difference of 1.04 cm.

关键词： Engineering   Public health   Computer engineering   Health sciences   Computer science   Medical imaging  

摘要： With the current covid-19 pandemic ongoing, staying safe is of utmost importance. In being proactive, people everywhere are clothed in facial masks and gloves. Businesses and universities are conducting temperature checks in a multitude of methods such as thermal imaging, forehead, oral, and aural thermometer handguns. From my research, I discovered that the non-contact-based devices available in the market today are quite expensive, ranging from $50-$60k (Juniper,[8]) and would still require some close contact with the next individual. The Artificial Intelligence and Visualizations (Ai&Vi) team at North Carolina A&T State University aimed to create fever detection systems that would promote safer testing regarding exposure to any viruses or illness one may carry, affordability, as well as equal or greater accuracy. For this research, I designed, built, and programmed three devices, namely: (i) Aggie Arm, a thermometer inspired device that reads temperature values from an individual’s forehead when within range; (ii) Aggie Eye, a thermal imaging device that measures body temperature based on values gathered from the tear duct of the eye with higher accuracy; (iii) Aggie Talk, an intercom system coupled with the Aggie Eye to be integrated with existing k-12 school intercom systems. Expanding on fever detection, I also wanted to implement the use of stereoscopic cameras to help improve our systems. The stereoscopic cameras will allow our devices to detect the presence of human interference as well as other objects that may block or interfere with temperature readings given by the thermal systems more accurately. The stereoscopic cameras will also calculate and display the depth measurement of the person or object from the stereo camera. This measurement can then be used for social distance monitoring, and improving temperature reading accuracies. Furthermore, we can apply the stereoscopic algorithm to an autonomous driving project where it would be an integral part of

关键词： Computer science   Artificial intelligence   Computer engineering  

摘要： In the past few decades, the advancement in technology and the internet has leveraged the application of social networks in the world. Millions of people connect through social networks irrespective of their geographical boundaries. These users tend to form communities on common ground, such as similar hobbies, school, work, and much more. Deep level mining and analysis of these communities, connections within these communities, and the connections within the users of these communities divulge abundant data about the underlying features of these complex networks. This thesis focuses on detecting communities at specific times to track the change in memberships. Doing so would allow one to observe how these communities evolved over the given period of time. For evaluation, four temporal social network datasets were considered from different backgrounds, such as a college network database, Facebook database, and two question-answer-comment website databases. Three different kinds of community detection methodologies such as Leiden, Louvain and Infomap are used to detect communities among these datasets. Two types of analysis are done on community data, structural level, and temporal level. The temporal level analysis is further conducted on the community level (both pairwise and individual) and node level. Specific metrics used in these kinds of analysis differ from each other in various ways. The resulting outcomes of this project may be used for predicting the memberships that may form eventually. Moreover, similar patterns may assist data analytics experts in detecting influential members, malicious activities in the network, a population of individuals suffering from specific health issues, and designing the recommendation systems.

关键词： computer science   curriculum   data science   education   graduate   interdisciplinary   statistics  

摘要： Organizing a graduate program in statistics and data science raises many questions, offering a variety of opportunities while presenting a multitude of choices. The call for graduate programs in statistics and data science is overwhelming. How does it align with other (future) study programs at the secondary and postsecondary levels? What could or should be the natural home for data science in academia? Who meets the entry criteria, and who does not? Which strategic choices inevitably play a prominent role when developing a curriculum? We share our views on the why, when, where, who and what.

关键词： Computer science   Artificial intelligence   Robotics  

关键词： Computer science   Bioinformatics   Applied mathematics   Biology   Epidemiology  

摘要： The recent COVID-19 pandemic and its management has highlighted a need for new and innovative approaches to modeling epidemics. This thesis seeks to address some of those needs by providing a new framework for modeling disease using a distributed, network based approach, dubbed NetworkSIR. It then illustrates the potentials of this model by extending it to allow for disease to spread through the environment rather than exclusively person-to-person contact. This is of note as all known network-based epidemic models consider only the case in which infection spreads through direct human-to-human contact, or host-to- human contact. As "superspreader" events often result from disease spreading through poor environmental conditions, EnvironmentalSIR, while simple, delivers a powerful advantage over traditional modeling techniques by being able to capture these events.

关键词： Energy   Algorithms   Soccer   Polyhedra   Computer science   Mathematics  

摘要： This thesis addresses two distinct subjects involving the point-generated Voronoi tessellations of 2-dimensional domains; in particular, the focus is on the 2-sphere and the square flat 2-torus where we dispense of *** one of the discussion examines the centroidal Voronoi energy from a global optimization perspective; with its applications rooted in vector quantization, numerical analysis and optimal location problems, we strive to elucidate the nature as well as approaching the ground state of the associated nonconvex minimization problem. The presentation starts by poking the reader's interest on the complex combinatorial structure of “low energy” configurations and by justifying the mathematical pertinence and richness of the quest for Voronoi energy global minima. To further tease on the subject, frameworks based on Bravais lattices and soccer ball topologies are showcased in order to educatedly try to approach the global minimizer on the square flat torus and the sphere respectively. On the same front a novel iterative coupling method known as MACN is introduced. Relying on alternating the celebrated Lloyd's algorithm with simple geometric-based initialization and annealing stages, this method allows to deterministically probe low energy basins of attraction and associated minima. Through a large statistical analysis encompassing the energy as well as other metrics, we empirically demonstrate that this method achieves far better results than the available state-of-the-art deterministic methods whilst retaining important performing advantages compared to stochastic alternatives. Overall, a strong empirical connection between the Voronoi energy and the isoperimetric ratio is drawn throughout this portion of the presentation. Finally, the second part of this work introduces an innovative first order (in time) model whose dynamics allow for the emergence of biological collective behavior. Characterized by a Voronoi topology perception framework (VTP), t