<div> <div> <div> <p>The objective of this study is that combination of molecular dynamics (MD) simulations and machine learning to complement each other. In this study, four steps are conducted. </p> <p>First is based on the empirical potentials development in silicon nanowires for theory parts of molecular dynamics. Many-body empirical potentials have been developed for the last three decades, and with the advance of supercomputers, these potentials are expected to be even more useful for the next three decades. Atomistic calculations using empirical potentials can be particularly useful in understanding the structural aspects of Si or Si-H systems, however, existing empirical potentials have many errors of parameters. We propose a novel technique to understand and construct interatomic potentials with an emphasis on parameter fitting, in which the relationship between material properties and potential parameters is explained. The input database has been obtained from density functional theory (DFT) calculations with the Vienna ab initio simulation package (VASP) using the projector augmented-wave method within the generalized gradient approximation. The DFT data are used in the fitting process to guarantee the compatibility within the context of multiscale modeling. </p> <p>Second, application part of MD simulations, enhancement of mechanical properties was focused in this research by using MEAM potentials. For instance, Young’s modulus, ultimate tensile strength, true strain, true stress and stress-strain relationship were calculated for nanosized Cu-precipitates using quenching & partitioning (Q&P) processing and nanosized Fe3C strengthened ultrafine-grained (UFG) ferritic steel. In the stress-strain relationship, the structure of simulation is defined using the constant total number of particles, constant-energy, constant-volume ensemble (NVE) is pulled in the y-direction, or perpendicular to the boundary interface, to increase strain. The strain in increased for a specified number of times in a loop and the stress is calculated at each point before the simulation loops.</p></div></div> </div> <div> <div> <div> <p>Third, based on the MD simulations, machine learning and the peridynamics are applied to prediction of disk damage patterns. The peridynamics is the nonlocal extension of classical continuum mechanics and same as MD model. Especially, FEM is based on the partial differential equations, however, partial derivatives do not exist on crack and damage surfaces. To complement this problem, the peridynamics was used which is based on the integral equations and overcome deficiencies in the modeling of deformation discontinuities. In this study, the forward problem (i), if we have images of damage and crack, crack patterns are predicted by using trained data compared to true solutions which are hit by changing the x and y hitting coordinates on the disk. The inverse problem (ii), if we have images of damage and crack, the corresponding hitting location, indenter velocity and indenter size are predicted by using trained data. Furthermore, we did the regression analysis for the images of the crack patterns with Neural processes to predict the crack patterns. In the regression problem, by representing the results of the variance according to the epochs, it can be confirmed that the result of the variance is decreased by increasing the epoch through the neural processes. Therefore, the result of the training gradually improves, and the ranges of the variance are expressed as 0 to 0.035. The most critical point of this study is that the neural processes makes an accurate prediction even if the information of the training data is missing or not enough. The results show that if the context points are set to 10, 100, 300, and 784, the training information is deliberately omitted such as context points of 10, 100 and 300, and the predictions are different when context points are significantly lower. However, when comparing the results of context points 100 and 784, the predicted results appear to be very similar to each other because of the Gaussian processes in the neural processes. Therefore, if the training data is trained through the Neural processes, the missing information of training data can be supplemented to predict the results. </p> <p>Finally, we predicted the data by applying various data using deep learning as well as MD simulation data. This study applied the deep learning to Cryo-EM images and Line Trip (LT) data with power systems. In this study, deep learning method was applied to reduce the effort of selection of high-quality particles. This study proposes a learning frame structure using deep learning and aims at freeing passively selecting high quality particles as the ultimate goal. For predicting the line trip data and bad data detection, we choose to analyze the frequency signal because suddenly the frequency changes in the power system due to events such as generator trip, line trip or load shedding in large power systems. </p> </div> </div> </div>

Prediction techniques based on data are applied in a broad range of applications such as bioinformatics, disease spread, and mobile intrusion detection, just to name a few. With the rapid emergence of on-line technologies numerous techniques for collecting and storing data for prediction-based analysis have been proposed in the literature. With the growing size of global population, the spread of epidemics is increasing at an alarming rate. Consequently, public and private health care officials are in a dire need of developing technological solutions for managing epidemics. Most of the existing syndromic surveillance and disease detection systems deal with a small portion of a real dataset. From the communication network perspective, the results reported in the literature generally deal with commonly known network topologies. Scalability of a disease detection system is a real challenge when it comes to modeling and predicting disease spread across a large population or large scale networks. In this dissertation, we address this challenge by proposing a hierarchical aggregation approach that classifies a dynamic disease spread phenomena at different scalability levels. Specifically, we present a finite state model (SEIR-FSM) for predicting disease spread, the model manifests itself into three different levels of data aggregation and accordingly makes prediction of disease spread at various scales. We present experimental results of this model for different disease spread behaviors on all levels of granularity. Subsequently, we present a mechanism for mapping the population interaction network model to a wireless mobile network topology. The objective is to analyze the phenomena of malware spread based on vulnerabilities. The goal is to develop and evaluate a wireless mobile intrusion detection system that uses a Hidden Markov model in connection with the FSM disease spread model (HMM-FSM). Subsequently, we propose a software-based architecture that acts as a network function virtualization (NFV) to combat malware spread in IoT based networks. Taking advantage of the NFV infrastructure's potential to provide new security solutions for IoT environments to combat malware attacks. We propose a scalable and generalized IDS that uses a Recurrent Neural Network Long Short Term Memory (RNN-LSTM) learning model for predicting malware attacks in a timely manner for the NFV to deploy the appropriate countermeasures. The analysis utilizes the susceptible (S), exposed (E), infected (I), and resistant (R) (SEIR) model to capture the dynamics of the spread of the malware attack and subsequently provide a patching mechanism for the network. Our analysis focuses primarily on the feasibility and the performance evaluation of the NFV RNN-LSTM proposed model.

Many tasks of visual computing and communications such as object recognition, matting, compression, etc., need to extract and encode the outer boundary of the object in a digital image or video. In this thesis, we focus on a particular video segmentation task and propose an efficient method for head-and-shoulder of humans through video frames. The key innovations for our work are as follows: (1) a novel head descriptor in polar coordinate is proposed, which can characterize intrinsic head object well and make it easy for computer to process, classify and recognize. (2) a learning-based method is proposed to provide highly precise and robust head-and-shoulder segmentation results in applications where the head-and-shoulder object in the question is a known prior and the background is too complex. The efficacy of our method is demonstrated on a number of challenging experiments. / Thesis / Master of Applied Science (MASc)

A case study for the tool LBTest illustrating benets and limitations of the tool along the terms of usability, results and costs. The study shows the use of learning based testing on a safety critical application in the avionics industry. While requiring the user to have the oretical knowledge of the tools inner workings, the process of using the tool has benefits in terms of requirement analysis and the possibility of finding design and implementation errors in both the early and late stages of development

Yes / Point set registration has been actively studied in computer vision and graphics. Optimization algorithms are at the core of solving registration problems. Traditional optimization approaches are mainly based on the gradient of objective functions. The derivation of objective functions makes it challenging to find optimal solutions for complex optimization models, especially for those applications where accuracy is critical. Learning-based optimization is a novel approach to address this problem, which learns the gradient direction from datasets. However, many learning-based optimization algorithms learn gradient directions via a single feature extracted from the dataset, which will cause the updating direction to be vulnerable to perturbations around the data, thus falling into a bad stationary point. This paper proposes the General Discriminative Optimization (GDO) method that updates a gradient path automatically through the trade-off among contributions of different features on updating gradients. We illustrate the benefits of GDO with tasks of 3D point set registrations and show that GDO outperforms the state-of-the-art registration methods in terms of accuracy and robustness to perturbations.

<div>Fluorescence microscopy has emerged as a powerful tool for studying cell biology because it enables the acquisition of 3D image volumes deeper into tissue and the imaging of complex subcellular structures. Fluorescence microscopy images are frequently distorted by motion resulting from animal respiration and heartbeat which complicates the quantitative analysis of biological structures needed to characterize the structure and constituency of tissue volumes. This thesis describes a two pronged approach to quantitative analysis consisting of non-rigid registration and deep convolutional neural network segmentation. The proposed image registration method is capable of correcting motion artifacts in three dimensional fluorescence microscopy images collected over time. In particular, our method uses 3D B-Spline based nonrigid registration using a coarse-to-fine strategy to register stacks of images collected at different time intervals and 4D rigid registration to register 3D volumes over time. The results show that the proposed method has the ability of correcting global motion artifacts of sample tissues in four dimensional space, thereby revealing the motility of individual cells in the tissue.</div><div><br></div><div>We describe in thesis nuclei segmentation methods using deep convolutional neural networks, data augmentation to generate training images of different shapes and contrasts, a refinement process combining segmentation results of horizontal, frontal, and sagittal planes in a volume, and a watershed technique to enumerate the nuclei. Our results indicate that compared to 3D ground truth data, our method can successfully segment and count 3D nuclei. Furthermore, a microscopy image synthesis method based on spatially constrained cycle-consistent adversarial networks is used to efficiently generate training data. A 3D modified U-Net network is trained with a combination of Dice loss and binary cross entropy metrics to achieve accurate nuclei segmentation. A multi-task U-Net is utilized to resolve overlapping nuclei. This method was found to achieve high accuracy object-based and voxel-based evaluations.</div>

In this thesis we investigate the feasibility of learning-based testing (LBT) as a viable testing methodology for reactive systems. In LBT, a large number of test cases are automatically generated from black-box requirements for the system under test (SUT) by combining an incremental learning algorithm with a model checking algorithm. The integration of the SUT with these algorithms in a feedback loop optimizes test generation using the results from previous outcomes. The verdict for each test case is also created automatically in LBT. To realize LBT practically, existing algorithms in the literature both for complete and incremental learning of finite automata were studied. However, limitations in these algorithms led us to design, verify and implement new incremental learning algorithms for DFA and Kripke structures. On the basis of these algorithms we implemented an LBT architecture in a practical tool called LBTest which was evaluated on pedagogical and industrial case studies. The results obtained from both types of case studies show that LBT is an effective methodology which discovers errors in reactive SUTs quickly and can be scaled to test industrial applications. We believe that this technology is easily transferrable to industrial users because of its high degree of automation. / <p>QC 20130312</p>

The purpose of this study is to evaluate the English teaching program at G&uuml / lhane Military Medical Faculty and suggest a new program based on the Monitor Model. The study, as an example of a systematic study of a language program and that of a proposed syllabus, is expected to aid the practice of English Language Teaching in Turkey. The data sources were 230 students, 25 doctors and 7 teachers. The data analysis showed that students do not like the contents of the course books. Students reported speaking and reading as priority skills. To understand and translate medical material, to get an overseas assignment, to talk to foreigners and to follow lectures were the common language-related goals. Students&amp / #8217 / , institution&amp / #8217 / s and doctors&amp / #8217 / needs and goals and available resources were surveyed and a new second language teaching program was suggested for Phase 1. A general curriculum model and a program design model were also suggested together with the syllabuses for Phase 1. In the suggested program, grammar, writing and other conscious learning activities are separated from comprehension or (subconscious) acquisition-based activities. The suggested design has three topic-based syllabuses organized in modular format for three levels: Advanced, intermediate and elementary. For the majority advanced level classes new materials need to be developed and for elementary and intermediate levels new course books are suggested. A sample module was prepared, piloted and the results are discussed. The piloted module was found better than the previous form of the lessons both by the students and the teachers.

Submitted by Alice Araujo (alice.caraujo@ufpe.br) on 2017-11-29T16:49:07Z No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) TeseFinal_fpms.pdf: 108609391 bytes, checksum: c84c50e3c8588d6c85e44f9ac6343200 (MD5) / Made available in DSpace on 2017-11-29T16:49:07Z (GMT). No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) TeseFinal_fpms.pdf: 108609391 bytes, checksum: c84c50e3c8588d6c85e44f9ac6343200 (MD5) Previous issue date: 2016-03-07 / CNPQ / Alignment of virtual elements to the real world scenes (known as detection and tracking) relying on features that are naturally present on the scene is one of the most important challenges in Augmented Reality. When it goes to complex scenes like industrial scenarios, the problem gets bigger with the lack of features and models, high specularity and others. Based on these problems, this PhD thesis addresses the question “How to improve object detection and pose estimation from natural features for AR when dealing with complex scenes problems?”. In order to answer this question, we need to ask ourselves “What are the challenges that we face when developing a new tracker for real world scenarios?”. We begin to answer these questions by developing a complete tracking system that tackles some characteristics typically found in industrial scenarios. This system was validated in a tracking competition organized by the most important AR conference in the world, called ISMAR. During the contest, two complementary problems to tracking were also discussed: calibration, procedure which puts the virtual information in the same coordinate system of the real world, and 3D reconstruction, which is responsible for creating 3D models of the scene to be used for tracking. Because many trackers need a pre-acquired model of the target objects, the quality of the generated geometric model of the objects influences the tracker, as observed on the tracking contest. Sometimes these models are available but in other cases their acquisition represents a great effort (manually) or cost (laser scanning). Because of this we decided to analyze how difficult it is today to automatically recover 3D geometry from complex 3D scenes by using only video. In our case, we considered an electrical substation as a complex 3D scene. Based on the acquired knowledge from previous experiments, we decided to first tackle the problem of improving the tracking for scenes where we can use recent RGB-D sensors during model generation and tracking. We developed a technique called DARP, Depth Assisted Rectification of Patches, which can improve matching by using rectified features based on patches normals. We analyzed this new technique under different synthetic and real scenes and improved the results over traditional texture based trackers like ORB, DAFT or SIFT. Since model generation is a difficult problem in complex scenes, our second proposed tracking approach does not depend on these geometric models and aims to track texture or textureless objects. We applied a supervised learning technique, called Gradient Boosting Trees (GBTs) to solve the tracking as a linear regression problem. We developed this technique by using image gradients and analyzing their relationship with tracking parameters. We also proposed an improvement over GBTs by using traditional tracking approaches together with them, like intensity or edge based features which turned their piecewise constant function to a more robust piecewise linear function. With the new approach, it was possible to track textureless objects like a black and white map for example. / O alinhamento de elementos virtuais com a cena real (definido como detecção e rastreamento) através de características naturalmente presentes em cena é um dos grandes desafios da Realidade Aumentada. Quando se trata de cenas complexas, como cenários industriais, o problema se torna maior com objetos pouco texturizados, alta especularidade e outros. Com base nesses problemas, esta tese de doutorado aborda a questão "Como melhorar a detecção de objetos e a estimativa da sua pose através de características naturais da cena para RA ao lidar com problemas de cenários complexos?". Para responder a essa pergunta, precisamos também nos perguntar: Quais são os desafios que enfrentamos ao desenvolver um novo rastreador para cenários reais?". Nesta tese, começamos a responder estas questões através da criação de um sistema de rastreamento completo que lida com algumas características tipicamente encontradas em cenários industriais. Este sistema foi validado em uma competição de rastreamento realizada na principal conferência de RA no mundo, chamada ISMAR. Durante a competição também foram discutidos dois problemas complementares ao rastreamento: a calibração, procedimento que coloca a informação virtual no mesmo sistema de coordenadas do mundo real, e a reconstrução 3D, responsável por criar modelos 3D da cena. Muitos rastreadores necessitam de modelos pré-adquiridos dos objetos presentes na cena e sua qualidade influencia o rastreador, como observado na competição de rastreamento. Às vezes, esses modelos estão disponíveis, mas em outros casos a sua aquisição representa um grande esforço (manual) ou custo (por varredura a laser). Devido a isto, decidimos analisar a dificuldade de reconstruir automaticamente a geometria de cenas 3D complexas usando apenas vídeo. No nosso caso, considerou-se uma subestação elétrica como exemplo de uma cena 3D complexa. Com base no conhecimento adquirido a partir das experiências anteriores, decidimos primeiro resolver o problema de melhorar o rastreamento para as cenas em que podemos utilizar sensores RGB-D durante a reconstrução e o rastreamento. Foi desenvolvida a técnica chamada DARP, sigla do inglês para Retificação de Patches Assistida por Informação de Profundidade, para melhorar o casamento de características usando patches retificados a partir das normais. A técnica foi analisada em cenários sintéticos e reais e melhorou resultados de rastreadores baseados em textura como ORB, DAFT ou SIFT. Já que a reconstrução do modelo 3D é um problema difícil em cenas complexas, a segunda abordagem de rastreamento não depende desses modelos geométricos e pretende rastrear objetos texturizados ou não. Nós aplicamos uma técnica de aprendizagem supervisionada, chamada Gradient Boosting Trees (GBTs) para tratar o rastreamento como um problema de regressão linear. A técnica foi desenvolvida utilizando gradientes da imagem e a análise de sua relação com os parâmetros de rastreamento. Foi também proposta uma melhoria em relação às GBTs através do uso de abordagens tradicionais de rastreamento em conjunto com a regressão linear, como rastreamento baseado em intensidade ou em arestas, propondo uma nova função de predição por partes lineares mais robusta que a função de predição por partes constantes. A nova abordagem permitiu o rastreamento de objetos não-texturizados como por exemplo um mapa em preto e branco.

<p><a>Internal ribosome entry sites (IRES) are segments of the mRNA found in untranslated regions, which can recruit the ribosome and initiate translation independently of the more widely used 5’ cap dependent translation initiation mechanism. IRES play an important role in conditions where has been 5’ cap dependent translation initiation blocked or repressed. They have been found to play important roles in viral infection, cellular apoptosis, and response to other external stimuli. It has been suggested that about 10% of mRNAs, both viral and cellular, can utilize IRES. But due to the limitations of IRES bicistronic assay, which is a gold standard for identifying IRES, relatively few IRES have been definitively described and functionally validated compared to the potential overall population. Viral and cellular IRES may be mechanistically different, but this is difficult to analyze because the mechanistic differences are still not very clearly defined. Identifying additional IRES is an important step towards better understanding IRES mechanisms. Development of a new bioinformatics tool that can accurately predict IRES from sequence would be a significant step forward in identifying IRES-based regulation, and in elucidating IRES mechanism. This dissertation systematically studies the features which can distinguish IRES from nonIRES sequences. Sequence features such as kmer words, and structural features such as predicted MFE of folding, Q_MFE, and sequence/structure triplets are evaluated as possible discriminative features. Those potential features incorporated into an IRES classifier based on XGBboost, a machine learning model, to classify novel sequences as belong to IRES or nonIRES groups. The XGBoost model performs better than previous predictors, with higher accuracy and lower computational time. The number of features in the model has been greatly reduced, compared to previous predictors, by adding global kmer and structural features. The trained XGBoost model has been implemented as the first high-throughput bioinformatics tool for IRES prediction, IRESpy. This website provides a public tool for all IRES researchers and can be used in other genomics applications such as gene annotation and analysis of differential gene expression.</a></p>