Physical Sciences and Mathematics Commons^™

Inexact Tensor Methods And Their Application To Stochastic Convex Optimization, Artem Agafonov, Dmitry Kamzolov, Pavel Dvurechensky, Alexander Gasnikov, Martin Takac

Machine Learning Faculty Publications

We propose general non-accelerated and accelerated tensor methods under inexact information on the derivatives of the objective, analyze their convergence rate. Further, we provide conditions for the inexactness in each derivative that is sufficient for each algorithm to achieve a desired accuracy. As a corollary, we propose stochastic tensor methods for convex optimization and obtain sufficient mini-batch sizes for each derivative. © 2020, CC BY.

A Deep Machine Learning Approach For Predicting Freeway Work Zone Delay Using Big Data, Abdullah Shabarek

Dissertations

The introduction of deep learning and big data analytics may significantly elevate the performance of traffic speed prediction. Work zones become one of the most critical factors causing congestion impact, which reduces the mobility as well as traffic safety. A comprehensive literature review on existing work zone delay prediction models (i.e., parametric, simulation and non-parametric models) is conducted in this research. The research shows the limitations of each model. Moreover, most previous modeling approaches did not consider user delay for connected freeways when predicting traffic speed under work zone conditions. This research proposes Deep Artificial Neural Network (Deep ANN) and …

Distributed Load Testing By Modeling And Simulating User Behavior, Chester Ira Parrott

LSU Doctoral Dissertations

Modern human-machine systems such as microservices rely upon agile engineering practices which require changes to be tested and released more frequently than classically engineered systems. A critical step in the testing of such systems is the generation of realistic workloads or load testing. Generated workload emulates the expected behaviors of users and machines within a system under test in order to find potentially unknown failure states. Typical testing tools rely on static testing artifacts to generate realistic workload conditions. Such artifacts can be cumbersome and costly to maintain; however, even model-based alternatives can prevent adaptation to changes in a system …

A Neat Approach To Malware Classification, Jason Do

Master's Projects

Current malware detection software often relies on machine learning, which is seen as an improvement over signature-based techniques. Problems with a machine learning based approach can arise when malware writers modify their code with the intent to evade detection. This leads to a cat and mouse situation where new models must constantly be trained to detect new malware variants. In this research, we experiment with genetic algorithms as a means of evolving machine learning models to detect malware. Genetic algorithms, which simulate natural selection, provide a way for models to adapt to continuous changes in a malware families, and thereby …

Pyxtal_Ff: A Python Library For Automated Force Field Generation, Howard Yanxon, David Zagaceta, Binh Tang, David S. Matteson, Qiang Zhu

Physics & Astronomy Faculty Research

We present PyXtal_FF—a package based on Python programming language—for developing machine learning potentials (MLPs). The aim of PyXtal_FF is to promote the application of atomistic simulations through providing several choices of atom-centered descriptors and machine learning regressions in one platform. Based on the given choice of descriptors (including the atom-centered symmetry functions, embedded atom density, SO4 bispectrum, and smooth SO3 power spectrum), PyXtal_FF can train MLPs with either generalized linear regression or neural network models, by simultaneously minimizing the errors of energy/forces/stress tensors in comparison with the data from ab-initio simulations. The trained MLP model from PyXtal_FF is interfaced with …

Signature Identification And Verification Systems: A Comparative Study On The Online And Offline Techniques, Nehal Hamdy Al-Banhawy, Heba Mohsen, Neveen I. Ghali Prof.

Future Computing and Informatics Journal

Handwritten signature identification and verification has become an active area of research in recent years. Handwritten signature identification systems are used for identifying the user among all users enrolled in the system while handwritten signature verification systems are used for authenticating a user by comparing a specific signature with his signature that is stored in the system. This paper presents a review for commonly used methods for preprocessing, feature extraction and classification techniques in signature identification and verification systems, in addition to a comparison between the systems implemented in the literature for identification techniques and verification techniques in online and …

Data: The Good, The Bad And The Ethical, John D. Kelleher, Filipe Cabral Pinto, Luis M. Cortesao

Articles

It is often the case with new technologies that it is very hard to predict their long-term impacts and as a result, although new technology may be beneficial in the short term, it can still cause problems in the longer term. This is what happened with oil by-products in different areas: the use of plastic as a disposable material did not take into account the hundreds of years necessary for its decomposition and its related long-term environmental damage. Data is said to be the new oil. The message to be conveyed is associated with its intrinsic value. But as in …

Lidar Object Detection Utilizing Existing Cnns For Smart Cities, Vinay Ponnaganti

Master's Projects

As governments and private companies alike race to achieve the vision of a smart city — where artificial intelligence (AI) technology is used to enable self-driving cars, cashier-less shopping experiences and connected home devices from thermostats to robot vacuum cleaners — advancements are being made in both software and hardware to enable increasingly real-time, accurate inference at the edge. One hardware solution adopted for this purpose is the LiDAR sensor, which utilizes infrared lasers to accurately detect and map its surroundings in 3D. On the software side, developers have turned to artificial neural networks to make predictions and recommendations with …

Detecting Deepfakes With Deep Learning, Eric C. Tjon

Master's Projects

Advances in generative models and manipulation techniques have given rise to digitally altered videos known as deepfakes. These videos are difficult to identify for both humans and machines. Typical detection methods exploit various imperfections in deepfake videos, such as inconsistent posing and visual artifacts. In this paper, we propose a pipeline with two distinct pathways for examining individual frames and video clips. The image pathway contains a novel architecture called Eff-YNet capable of both segmenting and detecting frames from deepfake videos. It consists of a U-Net with a classification branch and an EfficientNet B4 encoder. The video pathway implements a …

End-To-End Learning Utilizing Temporal Information For Vision- Based Autonomous Driving, Dapeng Guo

Master's Projects

End-to-End learning models trained with conditional imitation learning (CIL) have demonstrated their capabilities in driving autonomously in dynamic environments. The performance of such models however is limited as most of them fail to utilize the temporal information, which resides in a sequence of observations. In this work, we explore the use of temporal information with a recurrent network to improve driving performance. We propose a model that combines a pre-trained, deeper convolutional neural network to better capture image features with a long short-term memory network to better explore temporal information. Experimental results indicate that the proposed model achieves performance gain …

Multi-Agent Deep Reinforcement Learning For Walkers, Inhee Park

Master's Projects

This project was motivated by seeking an AI method towards Artificial General Intelligence (AGI), that is, more similar to learning behavior of human-beings. As of today, Deep Reinforcement Learning (DRL) is the most closer to the AGI compared to other machine learning methods. To better understand the DRL, we compares and contrasts to other related methods: Deep Learning, Dynamic Programming and Game Theory.

We apply one of state-of-art DRL algorithms, called Proximal Policy Op- timization (PPO) to the robot walkers locomotion, as a simple yet challenging environment, inherently continuous and high-dimensional state/action space.

The end goal of this project is …

Malware Classification With Gaussian Mixture Model-Hidden Markov Models, Jing Zhao

Master's Projects

Discrete hidden Markov models (HMM) are often applied to the malware detection and classification problems. However, the continuous analog of discrete HMMs, that is, Gaussian mixture model-HMMs (GMM-HMM), are rarely considered in the field of cybersecurity. In this study, we apply GMM-HMMs to the malware classification problem and we compare our results to those obtained using discrete HMMs. As features, we consider opcode sequences and entropy-based sequences. For our opcode features, GMM-HMMs produce results that are comparable to those obtained using discrete HMMs, whereas for our entropy-based features, GMM-HMMs generally improve on the classification results that we can attain with …

The Use Of Evidential Reasoning Model With Biomarkers In Pancreatic Cancer Prediction, Qianhui Fan

Master's Projects

In this project, an evidential reasoning model is built to amalgamate factors that could be used in early detection of pancreatic cancer. Our machine learning model outputs a probability of a given patient having prostate cancer based on various input variables. These variables include health history factors, such as smoking and medical history, technical artifacts, such as biopsy sequencing technology, and genomic biomarkers such as mutational, transcriptional and methylomic profiles, cfDNA, and copy number variation. The dataset used in this project is a part of The Cancer Genome Atlas (TCGA) project and was collected from the National Cancer Institute (NIH) …

Reasoning About User Feedback Under Identity Uncertainty In Knowledge Base Construction, Ariel Kobren

Doctoral Dissertations

Intelligent, automated systems that are intertwined with everyday life---such as Google Search and virtual assistants like Amazon’s Alexa or Apple’s Siri---are often powered in part by knowledge bases (KBs), i.e., structured data repositories of entities, their attributes, and the relationships among them. Despite a wealth of research focused on automated KB construction methods, KBs are inevitably imperfect, with errors stemming from various points in the construction pipeline. Making matters more challenging, new data is created daily and must be integrated with existing KBs so that they remain up-to-date. As the primary consumers of KBs, human users have tremendous potential to …

Understanding The Dynamic Visual World: From Motion To Semantics, Huaizu Jiang

Doctoral Dissertations

We live in a dynamic world, which is continuously in motion. Perceiving and interpreting the dynamic surroundings is an essential capability for an intelligent agent. Human beings have the remarkable capability to learn from limited data, with partial or little annotation, in sharp contrast to computational perception models that rely on large-scale, manually labeled data. Reliance on strongly supervised models with manually labeled data inherently prohibits us from modeling the dynamic visual world, as manual annotations are tedious, expensive, and not scalable, especially if we would like to solve multiple scene understanding tasks at the same time. Even worse, in …

Machine Learning Model Selection For Predicting Global Bathymetry, Nicholas P. Moran

University of New Orleans Theses and Dissertations

This work is concerned with the viability of Machine Learning (ML) in training models for predicting global bathymetry, and whether there is a best fit model for predicting that bathymetry. The desired result is an investigation of the ability for ML to be used in future prediction models and to experiment with multiple trained models to determine an optimum selection. Ocean features were aggregated from a set of external studies and placed into two minute spatial grids representing the earth's oceans. A set of regression models, classification models, and a novel classification model were then fit to this data and …

Malware Classification Using Lstms, Dennis Dang

Master's Projects

Signature and anomaly based detection have long been quintessential techniques used in malware detection. However, these techniques have become increasingly ineffective as malware becomes more complex. Researchers have therefore turned to deep learning to construct better performing models. In this project, we create four different long-short term memory (LSTM) models and train each model to classify malware by family type. Our data consists of opcodes extracted from malware executables. We employ techniques used in natural language processing (NLP) such as word embedding and bidirection LSTMs (biLSTM). We also use convolutional neural networks (CNN). We found that our model consisting of …

Quantifying Deepfake Detection Accuracy For A Variety Of Natural Settings, Pratikkumar Prajapati

Master's Projects

Deep fakes are videos generated from a starting video of a person where that person's face has been swapped for someone else's. In this report, we describe our work to develop general, deep learning-based models to classify Deep Fake content. Our first experiments involved simple Convolution Neural Network (CNN)-based models where we varied how individual frames from the source video were passed to the CNN. These simple models tended to give low accuracy scores for discriminating fake versus non-fake videos of less than 60%. We then developed three more sophisticated models: one based on choosing test frames, one based on …

Bioinformatics Metadata Extraction For Machine Learning Analysis, Zachary Tom

Master's Projects

Next generation sequencing (NGS) has revolutionized the biological sciences. Today, entire genomes can be rapidly sequenced, enabling advancements in personalized medicine, genetic diseases, and more. The National Center for Biotechnology Information (NCBI) hosts the Sequence Read Archive (SRA) containing vast amounts of valuable NGS data. Recently, research has shown that sequencing errors in conventional NGS workflows are key confounding factors for detecting mutations. Various steps such as sample handling and library preparation can introduce artifacts that affect the accuracy of calling rare mutations. Thus, there is a need for more insight into the exact relationship between various steps of the …

Research On Recovering Of Complex Networks Based On Boundary Nodes Of Giant Connected Component, Zhe Wang, Jianhua Li, Kang Dong

Journal of System Simulation

Abstract: Network recovery is an important way to solve the inevitable failure,and the reasonable recovery strategy can reduce the cost of resource and improve the network robustness.In order to study the dynamic behavior of recovery process and the relationship between recovery and network robustness,a Recovery Model of Boundary Nodes (RMBN) based on boundary of giant connected component is proposed,and two network Recovery strategies,Average Recovery of Boundary Nodes (ARBN) strategy and Priority Recovery of Boundary Nodes (PRBN) strategy are designed.The simulation results of different recovery strategies on three network models show that with the increase of recovery ratio,the …

Non-Cooperative Target Feature Point Cloud Registration Optimization Based On Icp Algorithm, Wei Liang, Muyao Xue, Huo Ju, Jinjie Zhang

Journal of System Simulation

Abstract: Aiming at the pose measurement caused by non-cooperative targets in visual measurement that cannot provide cooperation information,the ICP(Iterative Closest Point) algorithm is used to register the point cloud down-sampling data acquired at different times to complete the relative pose measurement of the target.The point cloud data of the target at the current moment is obtained using the structure from motion algorithm and the feature point matching algorithms are compared based on threshold matching and optical flow matching method.The extracted feature points are reconstructed by triangulation.The relative pose changes of the object at different times are calculated by using …

Frequency Regulation Signal Reduction Methods For Aluminum Smelters, Zejian Feng, Shengfei Li, Shouzhen Zhu, Zhiyun Li, Xiaomin Bai

Journal of System Simulation

Abstract: The motion delay of tap changer of aluminum smelter rectifier downgrades frequency response precision in following high frequency regulation signals.Aiming to improve the regulation performance of aluminum smelter loads based on adjusting dynamics,a fast regulation signal reduction technique is proposed,which is motion control threshold-based.A simulation-based decision technique of threshold values of the signal reduction algorithm's is devised to improve the performances of frequency response.Simulation results verify the efficacy of the techniques in promoting the frequency response precision to a certain production level,and contribute a practical way to support aluminum smelter‘s provision of frequency regulation services …

Research On Fuzzy Control And Optimization For Traffic Lights At Single Intersection, Jiajia Liu, Xingquan Zuo

Journal of System Simulation

Abstract: Aiming at the traffic signal control at urban single intersection,a fuzzy control method for traffic lights is presented.The method is based on a four-phase phasing sequence to control the traffic lights at a single intersection.Inputs of the fuzzy controller are the number of vehicles in line and the arrival rate of vehicles,and the output is the green light extension time of the current green light phase.A genetic algorithm (GA) is used to optimize fuzzy rules and membership functions of the fuzzy control system to improve the performance of the fuzzy controller.The fuzzy control method is realized by using …

Research On Dissemination And Control Of Public Opinion Based On Multilayer Coupled Network, Chen Shuai

Journal of System Simulation

Abstract: In order to study the influence of information interaction between multi-platforms on the dissemination and control of public opinion,taking Wechat and Weibo for example,a public opinion communication and control model based on multi-layer coupled network including Wechat layer,Weibo layer and control layer is constructed using multi-agent modeling method and improved SEIR model.On Anylogic platform,a simulation experiment was conducted on the event that “the use of materials by the Hubei Red Cross Society raises doubts”,and the effects factors such as single/dual platform,control range,control dynamics,control time and interaction between platforms were analyzed.The media guidance and government intervention strategies under multi-platform …

Amorphous Sio2/Si Interface Defects And Mechanism Of Passivation/Depassivation Reaction, Zhuocheng Hong, Zuo Xu

Journal of System Simulation

Abstract: The amorphous silicon dioxide/silicon (a-SiO₂/Si) interface is an important part of semiconductor devices.The passivation and depassivation process of silicon dangling bond defects (P_b-type defects) at the SiO₂/Si interface has a significant impact on semiconductor devices.Based on molecular dynamics and first-principles calculation methods,a-SiO₂/Si(111) interface model is constructed based on a-SiO₂ and crystalline Si.The CI-NEB (Climbing Image-Nudged Elastic Band) method is used to study the passivation and depassivation reactions of H₂ and H atoms of P_b defects at the a-SiO₂/Si(111) interface. The curves,barriers,and transition state structures of …

Collaborative Optimization Of Production And Energy Consumption In Flexible Workshop, Ding Yu, Wang Yan, Zhicheng Ji

Journal of System Simulation

Abstract: Considering the problem of the multi-objective constrained flexible job-shop,the NSGA-Ⅱalgorithm based on hybrid mutation operator is proposed.In view of NSGA-II algorithm being prone to premature convergence,poisson average and gaussian operators are introduced to improve the global and local optimization ability of the algorithm.The optimal scheme is selected from the set of pareto solutions by adopting the strategy of FAHP-IEVM,which is the combination of subjective and objective evaluation method. The modified algorithm is tested and compared by a series of ZDT test functions.The results show that the convergence and diversity of the revised algorithm are improved obviously.The effectiveness of …

Application Of Simulation Technology In Football Overall Attack Training, Xu Nuo, Shuanglong Liu, Fugao Jiang

Journal of System Simulation

Abstract: Simulation technology has a broad prospect in the field of football application.At present,football attack training is mainly based on audio-visual,experience and on-the-spot practice,but it lacks professional scene representation and systematic theoretical support. Visual C++ development platform and simulation programming method are used to virtualize the football training and reproduce the overall attack drill of football under different training scenes.The results show that the simulation technology can achieve more training situations. It can be used as an auxiliary tool for football overall attack training,enrich the training means,help players accurately understand the tactical system,and promote the scientific development of football training.

Optimization Algorithm For Planar Led Distribution And Connection, Fei Yue, Zhiqiang Gui, Yuyou Yao, Benzhu Xu, Liping Zheng

Journal of System Simulation

Abstract: The distribution and grouping of planar LED can be modeled as a multi constraint optimization problem.A novel algorithm based on the centroidal capacity-constrained power diagram for LED distribution is proposed,to achieve the goal of uniform illumination of planar LED.An optimization algorithm of LED combination and connection based on the greedy strategy is proposed to save materials.The experiment results show that the proposed algorithms are simple,effective in layout and grouping with rapid convergence,and can be used in practical applications.

Multi-Agent Behavior Simulation For Metro Station Passenger, Zequn Li, Fengting Yan, Zhicai Shi, Yumei Jian, Changhua Hua, Yongzhan Si, Xiang Yang

Journal of System Simulation

Abstract: Metro station is a typical public place with large crowd density.The characteristics of crowd behavior and the guidance based on the characteristics of crowd behavior can effectively train the crowd for emergency evacuation.Adopting the method of multi-agent and characteristics by measuring station building scene,analyzing the influence factors of passenger behavior characteristics,based on the passenger conformity rule,the single-agent passenger route choice behavior model is established.The multiple agents behavioral decision system in the virtual metro stations is established,the WebVR experiment is used to research the influencing factors of passenger herd behavior and decision-making behavior,which provides a theoretical basis for …

Parallel Finite Element Simulations On Radiation Damage Effects Of Lateral Pnp Bjts, Wang Qin, Zhaocan Ma, Hongliang Li, Linbo Zhang, Benzhuo Lu

Journal of System Simulation

Abstract: The Zlamal finite element discretization is applied in the drift-diffusion model for the simulations of semiconductor devices.Combined with the coupled ionization damage model,the ionization damage effects of lateral PNP (LPNP) bipolar junction transistors (BJT) are simulated.The model and algorithm are implemented based on the three-dimensional parallel adaptive finite element toolbox PHG (Parallel Hierarchical Grid).The phenomena of excess base current and current gain degradation in LPNP BJTs are successfully simulated via numerical calculation. A large-scale numerical experiment with 100 million elements and 1 024 MPI processes is carried out,demonstrating the good parallel scalability of the algorithm.