教学课程资源库 更多>>

关键词： Information processing   Neural networks   Artificial intelligence   Computer science   History of Oceania  

摘要： Information about the flow incident on an object is encoded in the observable characteristics of the structural response. Fluid-structure interactions such as the flapping of flags and swaying of trees are ubiquitous in the built environment and in nature. This thesis explores methods by which visual observations of such fluid-structure interactions can be used to make quantitative, instantaneous flow speed measurements. Videos capturing the responses of flags, flexible cantilevered cylinders, and trees were collected in laboratory and field experiments. Visual anemometry techniques were developed to infer the incident wind speeds from the observed structural responses, using both data-driven and physical modeling approaches. Results suggest that observations of common structures such as trees can be leveraged for quantitative flow speed *** capturing the responses of flags, flexible cantilevered cylinders, and trees were collected in laboratory and field experiments. Visual anemometry techniques were developed to infer the incident wind speeds from the observed structural responses. A data-driven approach employed deep learning model architectures to estimate wind speeds from raw video clips, demonstrating the feasibility of measuring wind speeds directly from visual observations of structural responses. Physical models were developed to relate incident flow speed to both mean structural deflections and to structural sway amplitudes, providing two possible signals that can be used to infer flow speeds from time-series measurements of the resulting structure kinematics. These physical models proved to generalize to various test structures through normalization with respect to the structure response under a reference wind condition. The primary contributions of this work are the successful demonstration of quantitative wind speed inference from visually observed structural responses, and the development of two physical models that can be generalized to

关键词： Computer science   Artificial intelligence  

摘要： An ontology is a vocabulary that provides a formal description of entities within a domain and their relationships with other entities. Along with basic schema information, it also captures information in the form of metadata about cardinality, restrictions, hierarchy, and semantic meaning. With the rapid growth of semantic (RDF) data on the web, many organizations like DBpedia, Earth Science Information Partners (ESIP) are publishing more and more data in RDF format. The ontology alignment task aims at linking two or more different ontologies from the same domain or different domains. It is a process of finding the semantic relationship between two or more ontological entities and/or instances. Information/data sharing among different systems is quite limited because of differences in data based on syntax, structures, and semantics. Ontology alignment is used to overcome the limitation of semantic interoperability of current vast distributed systems available on the Web. In spite of the availability of large hierarchical domain-specific datasets, automated ontology mapping is still a complex problem. Over the years, many techniques have been proposed for ontology instance alignment, schema alignment, and link discovery. Most of the available approaches require human intervention or work within a specific domain. The challenge involves representing an entity as a vector that encodes all context information of the entity such as hierarchical information, properties, constraints, etc. The ontological representation is rich in comparison with the regular data schema because of metadata about various properties, constraints, relationship to other entities within the domain, etc. While finding similarities between entities this metadata is often overlooked. The second challenge is that the comparison of two ontologies is an intense operation and highly depends on the domain and the language that the ontologies are expressed in. Most current methods require human interven

关键词： Artificial intelligence   Computer science  

关键词： Computer science   Information technology   Pharmaceutical sciences   Databases   Datasets   Prescription drugs   Trends   Servers  

摘要： The aim of this dissertation was to create a method of analyzing several secondary data sets by way of statistical analysis as a way of turning data into knowledge. To date, there have been several studies that use FDA Adverse Event Reporting System (FAERS) data as a foundation. That FAERS data is then combined with other quantitative data sets, but none have been combined in the same fashion of this study. The research question is an investigation into whether or not there is a correlation between FAERS data and Social Media mentions via Google Trends data. The drugs in the study were limited by being top-sellers in the United States for a minimum of 3 consecutive years. This study is of a quantitative correlational nature. The FAERS data was used as the dependent variable, and the Google Trends data was used as the independent variable. Based on the statistical analysis of using Pearson's Product-Moment Correlation, there were 6 different results that a drug can be categorized into. Results from this study could potentially be used by pharmaceutical companies, clinicians, or pharmacists to determine if post-regulatory approval marketing strategies need to be reevaluated for a given drug in order to help patients use it in a safer manner.

关键词： Computer science   Medical imaging  

关键词： Computer science   Mathematics   Epidemiology  

摘要： We present an individual-centric model for COVID-19 spread in an urban setting. We first analyze patient and route data of infected patients from January 20, 2020 ,to May 31, 2020, collected by the Korean Center for Disease Control & Prevention (KCDC) and illustrate how infection clusters develop as a function of time. This analysis offers a statistical characterization of mobility habits and patterns of individuals. We use this characterization to parameterize agent-based simulations that capture the spread of the disease, we evaluate simulation predictions with ground truth, and we evaluate different what-if counter-measure scenarios. Although the presented agent-based model is not a definitive model of how COVID-19 spreads in a population, its usefulness, limitations, and flexibility are illustrated and validated using hard data.

关键词： Robotics   Computer science   Electrical engineering   Artificial intelligence  

摘要： Autonomous robots are intelligent machines that perform given tasks without human interaction by interpreting information about their surroundings through various sensors. It is vital that such systems operate safely and be robust to various external changes. This requires accurate representation of the surroundings and robust estimation of its state from the environment perception sensors under measurement noise, environmental/external challenges, and sensor limitations. This dissertation focuses on novel approaches for building these robust and secure perception algorithms for robots that use cameras, LiDAR, and inertial sensors. Traditional Simultaneous localization and mapping (SLAM) systems estimate the motion of the robot and structure of the environment in the form of sparse or semi-dense maps using point features. The generated map is not helpful for tasks such as goal-based robot navigation, which require a semantic representation of the map, such as locations of the doors in an environment. Besides, the reliance of the SLAM system on geometry-based point features makes the system sensitive to viewpoint changes and various other environmental perturbations like light intensity changes that affect local scene appearance. Deep learning based semantic segmentation approaches can accurately label the environment using the images from the camera and be integrated with point feature-based SLAM pipelines. We propose a tightly coupled semantic RGB-D inertial odometry approach that estimates the robot's motion using camera, LiDAR, and inertial sensors by enhancing point features with semantic information and building a semantically labeled map of the environment. We show that our approach significantly reduces drift in a feature-sparse corridor environment, demonstrating that the SLAM odometry system and semantic understanding can benefit each other for robust perception. A key challenge towards robust perception is the fusion of data from multiple modalities with d

关键词： Computer science   Oncology   Artificial intelligence   COVID-19  

摘要： Today, there are only a few deep learning-based methods for computational drug design, none of which generates peptides. Our approach, GANDALF (Generative Adversarial Network Drug tArget Ligand Fructifier), solves this problem using deep GANS that include active atoms in the input data. Importantly, active atoms participate in covalent bonding between a target and ligand to facilitate binding. As a result, we can generate peptides for our targets in our diseases of interest: cancer and COVID-19. In this dissertation, we present both our prototype and final GANDALF methodologies and their results. We use casual convolutions and a modified policy gradient to help our GANs generate new peptides and structures. Casual convolutions ensure the model respects sequence order which allows us to generate reasonable peptide sequences and coordinates. Policy gradients use samples from our real data to reward the network for producing similar results. Our prototype method has some limitations despite using these techniques, such as generating fragmented peptide structures. For the final GANDALF, we present how we overcame these challenges using multiple GANs and supplementing our data for generating peptide structures using human-derived protein structures. We evaluate and compare our results using binding affinity, active atom distance, and sequence similarity. Binding affinity indicates the strength of chemical binding between a protein and ligand, giving us an indication of how well they interact. Active atom distances are the measured distances between the most active atom of the target and ligand after binding which helps us to evaluate how well they fit together in 3D space. We also consider sequence similarity to ensure our generated peptides are unique. In addition, we demonstrate the advantages of transfer learning and the contribution of active atoms in improving our generated peptides. The overall goal of this work is to generate new peptides for pharmaceutical drug d

关键词： Energy efficiency   Stochastic models   Algorithms   Feasibility studies   Traffic   Vehicles   Computer science   Energy   Transportation  

摘要： Difficulties that commercial vehicles are facing in meeting regulation standards require ad-hoc solutions. Emissions can be dramatically lowered if the characteristics of the transport application are known in advance. To tailor the vehicle's specification towards the use-case, however, a representative description of the mission, together with the surroundings, is needed. Where many conventional approaches fail, the operating cycle format (OC) has shown promising results in describing road operations in a way which is completely independent of both vehicle and *** specifically, the framework consists of three levels of representation. The first, called the bird's eye view, serves mainly as a classification tool, and makes use of metrics and labels to completely characterise the overall application of a vehicle during its lifetime. The second description, the stochastic operating cycle (sOC), condenses the main properties of a road operation using elementary statistics. It is conceived as an intermediate representation with a higher degree of accuracy. Finally, the deterministic operating cycle (dOC) is the most detailed description of a transport mission, and collects deterministic models to be used in simulation. In previous studies, the OC format was demonstrated to work in theory, but some margins for improvement could still be identified. Furthermore, the utility and benefits deriving from the use of the OC in concrete situations was explored only partially. The main objective of this thesis consists in extending the OC representation to include stochastic models for weather and traffic, which were missing in the original formulation. The new models are built to be parsimonious and to allow ease of parametrisation and implementation starting from real data. This enables to reproduce and simulate realistic environments where a transport mission can take place, with a substantial gain in accuracy. The second purpose of this work is to showcase how the OC

摘要： The article looks at several ways in which computer science topics can be made more accessible to students and librarians alike. They include the use of sophisticated unplugged computing activities such as game play or beadwork, doing book sorting activities and using an app that allows students to program their strategies, and learning block coding to create weavings.