教学课程资源库 更多>>

关键词： Aircraft   Turbines   Computers   Software utilities   Queuing   Design   Finite element analysis   Energy   Gas turbines   Distributed processing   Aerospace engineering   Computer science   Engineering  

摘要： The design of aero-derivative gas turbines (AGT) at Siemens Energy is a complex process that requires the use of various software tools for mechanical design, validation and testing. Due to the evolutionary nature of AGT development, changes are incremental and built upon previous models. Therefore, it is preferred to utilize the same tools that were initially developed and adopted decades ago in order to preserve the models and their real world offsets. Executing finite element analysis (FEA) at Siemens Energy is computationally intensive while its legacy tool is designed to be manually deployed onto local desktops and servers in a sequential manner. Consequently, the analysis of multiple AGT designs is a laborious process requiring users to invoke custom tool scripts on a limited pool of local computing resources. Users are responsible for maintaining each computing environment and have to checkpoint each analysis task manually. To address these challenges, I propose to cloudify the legacy FEA software tool by designing and developing an Analysis Tool as a Service (ATaaS) with a distributed microservice architecture that provides automation, scalability and distribution of the FEA software in a hybrid cloud deployment. The thesis contributions include (1) a distributed software architecture for ATaaS, (2) a prototype implementation for ATaaS integrating a legacy FEA tool used at Siemens and (3) various performance estimation models for executing FEA over ATaaS. (4) To validate the proposed architecture, models and prototype implementation, I carry out an experimental evaluation of performance and the functional testing of key components.

关键词： Computer science   Artificial intelligence   Electrical engineering  

关键词： Information Technology   Biomedical engineering   Computer science   Bioinformatics  

摘要： The digital transformation of healthcare over the past two decades has led to the proliferation of electronic health record (EHR) databases. These databases present an unprecedented opportunity for biomedical knowledge discovery. Data may be used for several purposes, including epidemiology, operational or clinical quality improvement studies, pragmatic trials and clinical trial recruitment, comparative effectiveness research, predictive modeling, clinical decision support, pharmacovigilance, and genome-wide association studies. In every case, one of the first steps involved is identifying the appropriate cohort of patients matching a set of inclusion and exclusion criteria, using only data available in the EHR. This process, known as EHR-driven phenotyping, is a resource-intensive task that involves many stakeholders, such as clinical experts, informaticists, and database analysts. It is therefore a critical rate-limiting factor that prevents massive scaling of knowledge discovery, and ultimately inhibits our ability to achieve the promise of national imperatives such as the Learning Healthcare System and All of Us. This research will attempt to improve the state of the art of EHR-driven phenotyping in three specific ways. First, we will analyze the variability of a set of existing, clinically validated, phenotype definitions in order to understand the requirements for a formal representation that supports automation. Second, we will assess the suitability of popular and emerging standards for formally representing cohort criteria, and evaluate whether this representation facilitates cross-platform cohort identification. Finally, we will develop and evaluate a fully standards-based system that can be used to create phenotype definitions and execute them against existing EHR data platforms, and evaluate the performance of this system in the context of the extant EHR-driven phenotyping ecosystem.

关键词： Computer science   Physiology   Neurosciences  

摘要： Sleep disorders are common among adults and children; it has serious consequences on their heath, cognitive development and quality of life. However, some sleep disorders are challenging to diagnose and more challenging to treat. Practitioners often rely on AIH for OSA patients’ classification task, where considering one measurement could raise a risk of oversimplification. Studies show the correlation between sleep disorders, specifically OSA, and mental health. On the other side there are an increasing number of studies suggested evidence of a relationship between the dynamic properties of functional brain structure with the behaviors and cognition attributes. This novel work objective is to assist medical professionals in screening and treating sleep disorders and any related cognitive disorders as a consequence. In this work we consider the novel problem of detecting behavioral measurements based on the resting-state fNIRS time series data. Taking the full advantages of the fNIRS neuroimaging technique, which is described as an emerging and promising technology for wide range of applications and as a diagnosing tool. In addition, it is a useful neuroimaging technology for research on children's cognitive development without the limitation of fMRI. The foundation of the study is capturing the spatiotemporal dynamic features of the resting-state FC and fully use the captured dynamic information for the prediction problem. To do so, we applied a 4D tensor method to build the dynamic functional connectivity matrices. Then the tensor statistical model trained on dFC matrices for the classification task. The highest classification accuracy achieved was 74%.

关键词： Mathematics   Computer science   Artificial intelligence   Search engines   Deep learning   Datasets   Inverse problems   Experiments   Corruption   Symmetry   Medical research   Dissertations & theses   Coronaviruses   COVID-19  

摘要： This dissertation serves as a collection of my three projects after I received the Ph.D. candidacy in 2018. The first two projects (in Chapter 2 and 3, respectively), joint works with Dongmian Zou and Gilad Lerman, are about novel algorithms for unsupervised and semi-supervised anomaly detection tasks, respectively. Our new methods allow datasets with a high ratio of corruption by outliers. The third project (in Chapter 4), a joint work with Kshitij Tayal, Raunak Manekar, Zhong Zhuang, Vipin Kumar and Ju Sun, brings out a methodology for improving the performance of end-to-end deep learning approaches for inverse problems with many-to-one forward mappings. General features of these three projects are introduced in the following. In Chapter 2, we propose a neural network for unsupervised anomaly detection with a novel robust subspace recovery layer (RSR layer). This layer seeks to extract the underlying subspace from a latent representation of the given data and removes outliers that lie away from this subspace. It is used within an autoencoder. The encoder maps the data into a latent space, from which the RSR layer extracts the subspace. The decoder then smoothly maps back the underlying subspace to a "manifold" close to the original inliers. Inliers and outliers are distinguished according to the distances between the original and mapped positions (small for inliers and large for outliers). Extensive numerical experiments with both image and document datasets demonstrate state-of-the-art precision and recall. In Chapter 3, we propose a new method for novelty detection that can tolerate high corruption of the training points, whereas previous works assumed either no or very low corruption. Our method trains a robust variational autoencoder (VAE), which aims to generate a model for the uncorrupted training points. To gain robustness to high corruption, we incorporate the following four changes to the common VAE: 1. Extracting crucial features of the latent code by a

关键词： Computer science   Physics   Geophysics   Archaeology   Remote sensing  

摘要： Deep learning models have been successfully applied to the task of object detection in photographic images. This success has led to experimentation with deep learning object detection in other types of image data, including subsurface radar images captured by ground-penetrating radar (GPR) devices. The high cost of GPR devices and of labeling GPR datasets has led to the use of synthetic GPR data for model training and evaluation in many prior studies. This work examines the current state of deep learning in the subsurface imaging field and expands upon it by evaluating the real-world performance of a synthetically-trained neural network model for object detection in GPR images. Further, three data augmentation techniques for GPR data are proposed and evaluated on a real-world test set: (1) random cropping, (2) negative augmentation, and (3) real noise application. The results of these data augmentation experiments indicate that none of the data augmentation techniques evaluated leads to improved real-world performance of a synthetically-trained GPR object detection model. The control experiment in which no data augmentation techniques were applied yielded the best-performing model as evaluated on the real-world test set. Additional work is required to determine which, if any, data augmentation techniques will improve the real- world performance of synthetically-trained GPR deep learning models. Several techniques are proposed as topics for future research. The results also demonstrate that a model that performs well on a synthetic validation set does not necessarily perform well on a real-world dataset. Several experiments produced models with an f-1 score of 0.90 or greater when evaluated on the synthetic validation set. These same models produced f-1 scores of 0.50 or lower (worse than random guessing) when evaluated on the real-world test set. The results demonstrate by counterexample that it is not possible to generalize about the real-world performance of a syn

关键词： Computer science   Finance   Economics   Fingers   Veins   Indexing   Forensics  

摘要： In second paragraph of the footnote on the first page of [1], the institution information of Lu Yang and Xiaoming Xi is inaccurate. The correct institution name is "School of Computer Science and Technology, Shandong University of Finance and Economics." So this paragraph should be corrected as:

关键词： Electrical engineering   Computer science   Biology  

摘要： Machine learning is a rapidly accelerating tool and technology used for countless applications in the modern world. There are many digital algorithms to deploy a machine learning program, but the most advanced and well-known algorithm is the artificial neural network (ANN). While ANNs demonstrate impressive reinforcement learning behaviors, they require large power consumption to operate. Therefore, an analog spiking neural network (SNN) implementing spike timing-dependent plasticity is proposed, developed, and tested to demonstrate equivalent learning abilities with fractional power consumption compared to its digital adversary.

关键词： Measurement   Feature extraction   Extraterrestrial measurements   Task analysis   Euclidean distance   Lighting   Computer science   Person re-identification   cross-modal retrieval   metric learning  

摘要： Person re-identification across visible and near-infrared cameras (VIS-NIR Re-ID) has widespread applications. The challenge of this task lies in heterogeneous image matching. Existing methods attempt to learn discriminative features via complex feature extraction strategies. Nevertheless, the distributions of visible and near-infrared features are disparate caused by modal gap, which significantly affects feature metric and makes the performance of the existing models poor. To address this problem, we propose a novel approach from the perspective of metric learning. We conduct metric learning on a well-designed angular space. Geometrically, features are mapped from the original space to the hypersphere manifold, which eliminates the variations of feature norm and concentrates on the angle between the feature and the target category. Specifically, we propose a cyclic projection network (CPN) that transforms features into an angle-related space while identity information is preserved. Furthermore, we proposed three kinds of loss functions, AICAL, LAL and DAL, in angular space for angular metric learning. Multiple experiments on two existing public datasets, SYSU-MM01 and RegDB, show that performance of our method greatly exceeds the SOTA performance.

关键词： Analytical chemistry   Atomic physics   Biology   Chemistry   Computer science   Medical imaging   Optics   Physics  

摘要： The adaptable surface chemistry of cellulose nanocrystals (CNCs) makes them outstanding; it provides colloidal stability, which is essential for engineering use, to commercially-available CNCs. Colloidal stability is achieved by the bulky negatively-charged sulfate half-ester groups on the surface that it manufactures itself via the use of sulfuric acid. The surface chemistry of CNCs can also be adapted into dialdehyde cellulose nanocrystals (DAC) by periodate oxidation in the presence of the sulfate half-ester groups. The oxidation extends the range of applications of CNCs. The objective of this thesis is to elucidate the role played by the sulfate half-ester groups on the oxidation reaction, both in the kinetics and its products. The results demonstrate that the oxidation reaction reduces the number of sulfate half-ester groups, which impacts on the colloidal stability, the size of the resulting product properties and, consequently, the DAC applications. A rheology study and a proof-of-concept demonstration are also performed to verify whether CNC derivatives could reinforce emulsions solidified by gelatin, as this could extend the range of CNCs in the field of microwave technology to be used as a fat phantom. The gelatin emulsions reinforced by CNC derivatives maintain the storage modulus above the loss modulus at temperatures above 40 ºC. The thermal stability of the reinforced emulsions could permit the successful implementation of these gels in the field of microwave technology. CNCs are possibly the better candidate than DAC for the reinforcement due to a combination of the ease of the phantom production and performance.