教学课程资源库 更多>>

关键词： Computer science   Artificial intelligence   Information Technology  

关键词： Archaeology   Computer science   Information technology  

摘要： Archaeological studies of maize (Zea mays) in the southeastern United States focus on its role in the rise of complex societies, the diet, and its economic importance (Fritz 2019; Scarry 1986; VanDerwarker et al. 2017). Many archaeologists recognize the relationship between maize agriculture and its role in the development of Mississippian societies (Anderson and Sassaman 2012; Cobb 2003; Fritz 2019; Scarry 1986; Welch and Scarry 1995). There was a widespread shift to maize agriculture in the North American Southeast that began around AD 900, reaching the Carolina Piedmont around AD 1000 (Gremillion 1989, 1993c; Hutchinson et al. 1998; Smith 1989). While the shift to maize cultivation has gained considerable attention, changes in food consumption—both food types and preparation—and how they relate to the vessels used to cook those foods have received less attention. In this thesis, I propose that, during the Late Woodland period (AD 800 to 1600) in the North Carolina Piedmont, cooking jar forms changed in response to the intensified cultivation of maize and the new foodways associated with it. Maize kernels require processing to make them safe for sustained human consumption, and I argue this new foodway, among other cuisine changes, and their processing requirements are what led to a transition in cooking jar forms from the Uwharrie phase (AD 800-1200) into the Dan River phase (AD 1000-1450). Through the frameworks of communities of practice (Wenger 1998) and communities of consumption (Mills 2016) I explain behaviors associated with pottery production and cooking practices and how the two are linked. In this case, changes in pottery practices are spurred by changes in diet. I use DesignCAD 3D MAX software to build models of Woodland period cooking jars which I analyze alongside use-wear evidence to determine the functions and methods of cooking used with them. I do this by graphing changes in height, rim diameter, volume, and neck constriction over time. Trends of

关键词： Computer science   Artificial intelligence  

关键词： Analytical chemistry   Chemistry   Computer science   Electromagnetics   Medical imaging   Morphology   Neurosciences   Optics   Physics  

摘要： Background and Purpose Chemical exchange saturation transfer (CEST) is a magnetic resonance imaging (MRI) technique, which employs a frequency selective preparation of the magnetization (called saturation) together with the chemical exchange of protons between metabolites and water. This process leads to a change of the observed MRI signal resulting in the generation of a contrast, which is dependent on the molecular environment of the tissue. CEST MRI can be performed at clinical field strengths of 3 T. Due to intrinsic properties of the process, quantitative CEST MRI measurements largely benefit from ultra-high magnetic fields (UHF) of 7 T and higher. However, at UHF the physical properties of CEST MRI and the imaging process result in the necessity of using additional correction methods to be able to achieve fully quantitative CEST acquisitions. These corrections, especially of the transmit field (B1+), result in a major increase of the acquisition time. The main goal of this thesis is to mitigate the influence of the B1+ field inhomogeneity on CEST MRI in order to reduce the required acquisition time of CEST MRI at UHF. Furthermore, the repeatability and reproducibility of CEST MRI in brain acquisitions after such a mitigation should be investigated. Methods In order to manipulate the spatial distribution of the B1+ field inhomogeneity a parallel transmit (pTx) system was used. Using this system, a new method called Multiple Interleaved Mode Saturation (MIMOSA) was designed and implemented. MIMOSA is based on the idea of time interleaving of two complimentary spatial distributions of the B1+ field in order to achieve a more homogeneous averaged effective field. Additionally, for 3D whole-brain acquisitions, MIMOSA was supplemented with a pTx excitation method called fast online-customized (FOCUS) pulses resulting in the development of the MIMOSA FOCUS sequence. Simulations using numerical solutions of the Bloch-McConnell equation were performed in order to evalu

关键词： Electrical engineering   Histology   Optics   Computer science  

关键词： Biochemistry   Bioinformatics   Chemistry   Computer science   Genetics   Systematic biology  

摘要： In the past few decades, advanced sequencing technologies and the availability of genomic sequences has made genome-wide studies possible. Nonetheless, understanding the structure and function of genomic elements, the so-called genome annotation, remains a major challenge. Comparative genomics plays an essential role in accurate genome annotation and understanding various genomic elements such as Transposable Elements (TE) and pseudogenes, among others. The impact of these dynamic components on genome structure and evolution as well as their contributions to the formation of other functional elements has created growing interest and demand for developing new annotation *** research is motivated by the fact that investigating ancestral genomes may shed some light on a more comprehensive understanding of our own genome. We developed bioinformatics approaches to identify very ancient TEs (Chapter 2) and pseudogenes (Chapter 3) that have shaped the dynamic components of the human genome, by making use of computationally reconstructed ancestral DNA sequences. These pipelines yield a large improvement over existing approaches and identify ancient genomic elements that have contributed significantly to the creation of new functional elements in the human *** annotation pipelines rely on computationally inferred ancestral genomes. Whole-genome multiple sequence alignment (MSA) of modern genomic sequences is a key to the reconstruction of ancestral sequences. Inaccurate reconstruction, as well as many of the inaccurate annotations in comparative genomics is rooted in MSA errors. In Chapter 4, we introduce an approach to improve MSA using our knowledge of TE consensus sequences. Improving the MSA accuracy not only helps us enhance our predictions but it is beneficial to other genomic projects involving MSA.

关键词： Computer science   Artificial intelligence   Automotive engineering  

摘要： Autonomous Driving (AD) systems are being researched and developed actively in recent days to solve the task of controlling the vehicles safely without human intervention. One method to solve such task is through deep Reinforcement Learning (RL) approach. In deep RL, the main objective is to find an optimal control behavior, often called policy performed by an agent, which is AD system in this case. This policy is usually learned through Deep Neural Networks (DNNs) based on the observations that the agent perceives along with rewards feedback received from environment. However, recent studies demonstrated the vulnerability of such control policies learned through deep RL against adversarial attacks. This raises concerns about the application of such policies to risk-sensitive tasks like AD. Previous adversarial attacks assume that the threats can be broadly realized in two ways: First one is targeted attacks through manipulation of the agent’s complete observation in real time and the other is untargeted attacks through manipulation of objects in environment. The former assumes full access to the agent’s observations at almost all time, while the latter has no control over outcomes of attack. This research investigates the feasibility of targeted attacks through physical adversarial objects in the environment, a threat that combines the effectiveness and practicality. Through simulations on one of the popular AD systems, it is demonstrated that a fixed optimal policy can be malfunctioned over time by an attacker e.g., performing an unintended self-parking, when an adversarial object is present. The proposed approach is formulated in such a way that the attacker can learn a dynamics of the environment and also utilizes common knowledge of agent’s dynamics to realize the attack. Further, several experiments are conducted to show the effectiveness of the proposed attack on different driving scenarios empirically. Lastly, this work also studies robustness of object loca

关键词： Electrical engineering   Medical imaging   Artificial intelligence   Computer science  

摘要： This thesis aims to develop an efficient, deep network based method for Magnetic Resonance Imaging (MRI) acceleration through undersampled MR image reconstruction. Deep Neural Networks, particularly Deep Convolutional Networks, have been demonstrated to be highly effective in a wide variety of computer vision tasks, including MRI reconstruction. However, modern highly efficient encoder structures, such as the EfficientNet can potentially reduce reconstruction times further while improving reconstruction quality. To that end, we have developed a multi-channel U-Net MRI reconstruction network which uses an EfficientNet encoder and a custom asymmetric. The network was trained and tested using 5x undersampled multi-channel brain MR image data from the Calgary Campinas dataset and was found to outperform comparable traditional U-Net structures in terms of image quality metric analysis and basic visual comparison while achieving a four-fold reduction in inference time.

关键词： Receivers & amplifiers   Aircraft   Kinematics   Real time   Sensors   Unmanned aerial vehicles   Algorithms   Satellites   Aerospace engineering   Computer science   Electrical engineering   Engineering   Mechanics   Physics   Robotics   Transportation  

摘要： In this thesis two methods were developed for precise maritime landing of an autonomous multirotor aircraft based on real-time kinematic (RTK) Global Navigation Satellite System (GNSS). The first method called RTK-localized method (RLM) uses RTK GNSS measurements to localize a sea vessel and execute the landing. RLM was demonstrated outdoors in hardware and landed on a physically simulated boat called a mock-boat with an average landing error of 9.7 cm. The mock-boat was actuated to have boat-like motion and a forward velocity of ~2 m/s. This method showed that accurate landing is possible with RTK GNSS as the primary means of localizing a sea vessel. The localization was unaided by non-GNSS sensors or an estimator, but lacked full attitude estimation and measurement smoothing. The second method was called RTK-Estimation Method (REM) and provides a more complete and robust solution, particularly at sea. It includes a base (landing pad) estimator to fuse RTK GNSS measurements with a dynamic model of a sea vessel. In contrast to RLM, the estimator provides full attitude estimation and measurement smoothing. The base estimator consists of an EKF in conjunction with a complimentary filter and estimates the relative position, attitude, and velocity of a moving target using RTK GNSS and inertial measurements alone. REM was demonstrated outdoors in hardware for 18 flight tests. The same mock-boat from RLM was used as a substitute for a sea vessel, and the boat motion varied between tests. These dynamics were recorded and performances were compared. The rate of success was high given moderate mock-boat motion and degraded with more aggressive motion. Tests were conducted with forward velocities from 0 to 3 m/s and moderate to high wave like motion. Over all tests for REM, the multirotor landed with an average accuracy of 12.7 cm. The methods described depart from common methods given that the only sensors involved for tracking the sea vessel were RTK GNSS receivers and iner

关键词： Autonomous underwater vehicles   Robots   Global positioning systems--GPS   Mathematical functions   Unmanned aerial vehicles   Algorithms   Localization   Sensors   Aerospace engineering   Computer science   Engineering   Mathematics   Robotics   Transportation  

摘要： In the real world, a robotic system must operate in the presence of motion and sensing uncertainty. This is caused by the fact that the motion of a robotic system is stochastic due to disturbances from the environment, and the states are only partially observable due noise in the sensor measurements. As a result, the true state of a robotic system is unknown, and estimation techniques must be used to infer the states from the belief, which is the probability distribution over all possible states. Accordingly, a robotic system must be capable of reasoning about the quality of the belief at future time steps to manage the growth of uncertainty by choosing the correct actions. This is problem is referred to as active localization and is the problem addressed in this *** goal of active localization is to plan a sensor trajectory that minimizes the growth uncertainty potentially along with additional objectives This problem can be formulated as a Partially Observable Markov Decision Process (POMDP), which is a mathematical framework for modelling sequential decision making problems inthe presence of uncertainty. However, obtaining an exact solutionto a POMDP is an intractable problem in general, and only a few problems formulated as POMDPs can be solved exactly. While significant progress has been made in recent years towards approximately solving POMDPs, existing methods suffer from suboptimality for problems where the objective is to reach a particular state. This dissertation considers approximate POMDP solutions and seeks to quantify the utility for active localization under various assumptions This dissertation can be separated in three main parts. First, uncertainty metrics (such as the A-, E-, and D-optimality criteria) are analyzed, which are necessary for quantifying uncertainty associated with the belief. A metric is proposed extending the criteria that allows for efficiently quantifying uncertainty at future time steps. Second, the metrics are ana