教学课程资源库 更多>>

关键词： Musical composition   Applied mathematics   Computer science  

摘要： The Material Point Method (MPM) has shown its high potential for physics-based simulation in the area of computer graphics. In this dissertation, we introduce a couple of improvements to the traditional MPM for different applications and demonstrate the advantages of our methods over the previous methods. First, we present a generalized transfer scheme for the hybrid Eulerian/Lagrangian method: the Polynomial Particle-In-Cell Method (PolyPIC). PolyPIC improves kinetic energy conservation during transfers, which leads to better vorticity resolution in fluid simulations and less numerical damping in elastoplasticity simulations. Our transfers are designed to select particle-wise polynomial approximations to the grid velocity that are optimal in the local mass-weighted L2 norm. Indeed our notion of transfers reproduces the original Particle-In-Cell Method (PIC) and recent Affine Particle-In-Cell Method (APIC). Furthermore, we derive a polynomial basis that is mass orthogonal to facilitate the rapid solution of the optimality condition. Our method applies to both of the collocated and staggered grid. As the second contribution, we present a novel method for the simulation of thin shells with frictional contact using a combination of MPM and subdivision finite elements. The shell kinematics are assumed to follow a continuum shell model which is decomposed into a Kirchhoff-Love motion that rotates the mid-surface normals followed by shearing and compression/extension of the material along the mid-surface normal. We use this decomposition to design an elastoplastic constitutive model to resolve frictional contact by decoupling resistance to contact and shearing from the bending resistance components of stress. We show that by resolving frictional contact with a continuum approach, our hybrid Lagrangian/Eulerian approach is capable of simulating challenging shell contact scenarios with hundreds of thousands to millions of degrees of freedom. Without the need for collision d

ISBN: (纸本)0822946203;9780822946205

关键词： Information science   Computer science  

摘要： Constant innovation in attack methods presents a significant problem for the security community which struggles to remain current in attack prevention, detection and response. The practice of threat hunting provides a proactive approach to identify and mitigate attacks in real-time before the attackers complete their objective. In this research, I present a matrix of adversary techniques inspired by MITRE’s ATT&CK matrix. This study allows threat hunters to classify the actions of advanced persistent threats (APTs) according to network-based behaviors.

关键词： Chemical engineering   Computer science  

摘要： In chemical manufacturing plants, numerous types of data are accessible, which could be process operational data (historical or real-time), process design and product quality data, economic and environmental (including process safety, waste emission and health impact) data. Effective knowledge extraction from raw data has always been a very challenging task, especially the data needed for a type of study is huge. Other characteristics of process data such as noise, dynamics, and highly correlated process parameters make this more challenging. In this study, we introduce an attention-based RNN for multi-step-ahead prediction that can have applications in model predictive control, fault diagnosis, etc. This model consists of an RNN that encodes a sequence of input time series data into a new representation (called context vector) and another RNN that decodes the representation into output target sequence. An attention model integrated to the encoder-decoder RNN model allows the network to focus on parts of the input sequence that are relevant to predicting the target sequence. The attention model is jointly trained with all other components of the model. By having a deep architecture, the model can learn a very complex dynamic system, and it is robust to noise. In order to show the effectiveness of the proposed approach, we perform a comparative study on the problem of catalyst activity prediction, by using conventional machine learning techniques such as Support Vector Regression (SVR).

关键词： Bioinformatics   Artificial intelligence   Computer science  

摘要： The development of theoretical computational methods and their application has become widespread in the world today. In this dissertation, I present my work in the creation of models to detect and describe complex biological and health related problems. The first major part of my work centers around the creation and enhancement of methods to calculate protein structure and dynamics. To this end, substantial enhancement has been made to the software package REDCRAFT to better facilitate its usage in protein structure calculation. The enhancements have led to an overall increase in its ability to characterize proteins under difficult conditions such as high noise and low data density. Secondly, a database that allows for easy and comprehensive mining of protein structures has been created and deployed. We show preliminary results for its application to protein structure calculation. This database, among other applications, can be used to create input sets for computational models for prediction of protein structure. Lastly, I present my work on the creation of a theoretical model to describe discrete state protein dynamics. The results of this work can be used to describe many real-world dynamic systems. The second major part of my work centers around the application of machine learning techniques to create a system for the automated detection of smoking using accelerometer data from smartwatches. The first aspect of this work that will be presented is binary detection of smoking puffs. This model was then expanded to perform full cigarette session detection. Next, the model was reformulated to perform quantification of smoking (such as puff duration and the time between puffs). Lastly, a rotational matrix was derived to resolve ambiguities of smartwatches due to position of the watch on the wrist.

关键词： Robotics   Artificial intelligence   Computer science  

摘要： Motion estimation using visual data is an essential aspect of today's self-driving cars, autonomous drones, robotics, and many other similar applications. Most proposed algorithms use either feature-based methods, which look at special features in images; or, direct methods, which look at the whole image pixels. Feature-based algorithms are mostly used with traditional descriptors, such as SIFT, SURF, Hessian, or Harris, to extract point-wise features and then match those features across the image frames in the camera's video input. Recently, local affine features (LAFs) have become a hot topic in many 3D vision applications, including relative pose estimation and Structure from Motion (SfM). This is because they are more informative than point features. The clear benefit to using affine correspondences (ACs) is the fact that they add three linear constraints on the Fundamental matrix F, hence fewer correspondences are required. This means the iterative, random sampling process of RANSAC will take fewer iterations to find the inlier matches. Therefore, the use of ACs improves performance by reducing the computation time, while maintaining similar accuracy. In this work, we focus on making motion estimation with affine correspondences more accurate, robust, and applicable to a wider range of setups. We also extend the epipolar constraint on affine correspondences to the multi-camera setting. Then, we propose a new solution for multi-camera motion estimation with affine correspondences using this constraint. Our new affine-based solver outperforms its point-based counterpart while maintaining similar accuracy. A statistical analysis of the experimental results on synthetic and real data confirms this conclusion. However, affine correspondences are known to be more vulnerable to noise than point correspondences. Current photometric refinement methods are costly; making such solutions unsuitable for real-time applications. We propose a new method to correct affine corre

关键词： Computer engineering   Artificial intelligence   Computer science  

摘要： This article presents two methods, REVAMPT and CARPe Posterum. REVAMPT, or Real-time Edge Video Analytics for Multi-person Privacy-aware Tracking, is an integrated system for privacy-built-in pedestrian re-identification. REVAMPT presents novel algorithmic and system constructs to push deep learning capabilities for pedestrian re-identification at the edge (i.e. the video camera). On the algorithm side, REVAMPT proposes a unified computer vision pipeline for detection and re-identification on a low-power computing device without the need for storing the streaming data. At the same time, it avoids facial recognition, re-identifying pedestrians based on their encoded key features at runtime. For the results and evaluation, this article also proposes a new metric, Accuracy·Efficiency (Æ), for holistic evaluation of deployable systems based on accuracy, performance, and power efficiency. REVAMPT outperforms current state-of-the-art by as much as ten-fold Æ improvement. A symbiotic task for pedestrian re-identification is path prediction, and therefore we also propose CARPe Posterum, a Convolutional Approach for Real-time Pedestrian Path Prediction. Having insight into the movement of pedestrians is not only important for pedestrian re-identification, but also for ensuring safe operation in a variety of applications including autonomous vehicles and social robotics. Current works in this area utilize complex generative or recurrent methods to capture many possible futures. However, despite the inherent real-time nature of predicting future paths, little work has been done to explore accurate and computationally efficient approaches for this task. To this end, CARPe utilizes a variation of Graph Isomorphism Networks in combination with an agile convolutional neural network design to form a fast and accurate path prediction approach. Notable results in both inference speed and prediction accuracy are achieved, improving FPS by at least 8x in comparison to current state-of-

关键词： Information science   Computer science  

摘要： The aim of this project is to identify, analyze, and rank local government websites based on their emergency informational response and preparedness for disasters. This exploratory research examines disaster information from multiple local city, town, and county government websites. Following the machine learning text analysis procedure of emergency management related information, the completed analysis for 1,611 local government websites represented a surviving legacy rate of 53% from the initial 3,000 crawled websites. All internal and external links of each government site were collected within the secondary uniform resource locator (URL) depth. This work extracts existing semantics associated with emergency topics to identify existing informational gaps between the local government and its citizens with respective measures of informational content cross-entropy and Kullback distance. Data mining was coded in R language. The statistical analysis was performed in SPSS. Spatial analysis was reviewed with Tableau and QGIS. The natural language processing was computed. The tool was tested on a 36 CPU station with 72 gigabyte memory. High and low scoring sites were identified with emergency phase representation. To serve these tasks, an analytic tool was developed within a streaming architecture. This allows the user to easily add new sitemap data to investigate topic patterns while performing a competitive analysis on a large volume of website links and structure analysis.

关键词： Biophysics   Computer science  

摘要： Studies of single biomolecules provide information that is buried in an ensemble-based measurement, including the evolution of an individual biomolecule’s behavior over time. Recent work showed that an electronic sensor composed of single-walled carbon nanotube field-effect transistors (SWCNT-FETs) can observe an individual biomolecule’s conformational motions over time and obtain accurate measurements of catalytic rates for a variety of enzymes. This dissertation expands the scope of transistor-based biosensing techniques through several strategies. The first strategy extends previous work by investigating similar enzymes, such as other DNA polymerases, in order to identify unique characteristics of each enzyme. A second strategy focuses on investigating the dynamics of biomolecular interactions that have not been previously studied by this technique, such as ligand-binding interactions. A third strategy makes refinements to the measurement or analysis techniques to uncover additional, subtler dynamics and other information that was previously hidden in the acquired signal. The dissertation is organized into two main parts. The first part (Chapters 1−3) discusses new measurements performed using the SWCNT-FET technique. Chapter 1 provides a brief introduction to the SWCNT-FET biosensing technique and details the methods and materials used in the experiments described in the following two chapters. Chapter 2 studies the behavior of a weakly-interacting antibody-antigen system: antibody 3C6 in the presence of the antigen paclitaxel. SWCNT-FET recordings of antigen-antibody binding exhibited two conductance states corresponding to bound and unbound configurations, like the two-level dynamics previously recorded from enzymatic catalysis. The SWCNT-FET signal correlated with antigen concentration, remaining relatively static at concentrations far from the value of the dissociation constant KD and fluctuating most actively near KD. Analysis of the distribution of single-

关键词： Information science   Computer science  

摘要： The purpose of this research is to identify how Machine Learning aids in Computer Forensics investigations. Semi-automated Machine Learning-enabled behavioral profiling techniques applied to information gathered from social media, personal computer data, and phone records increase the accuracy of forensic analysis and decrease time spent analyzing the information gathered from manual methods. ML provides these benefits flexibly, as it has the capability to be done by supervised or semi-supervised methods. Either method used results in the acquisition of more of the data needed for ML analysis techniques, and when coupled with a human analyst provides fast and accurate analytic conclusions by minimizing the impact of human nature and human faults on an investigation.