Physical Sciences and Mathematics Commons^™

Nonparametric Bayesian Deep Learning For Scientific Data Analysis, Devanshu Agrawal

Doctoral Dissertations

Deep learning (DL) has emerged as the leading paradigm for predictive modeling in a variety of domains, especially those involving large volumes of high-dimensional spatio-temporal data such as images and text. With the rise of big data in scientific and engineering problems, there is now considerable interest in the research and development of DL for scientific applications. The scientific domain, however, poses unique challenges for DL, including special emphasis on interpretability and robustness. In particular, a priority of the Department of Energy (DOE) is the research and development of probabilistic ML methods that are robust to overfitting and offer reliable …

Hierarchical Multimodal Attention For End-To-End Audio-Visual Scene-Aware Dialogue Response Generation, Hung Le, Doyen Sahoo, Nancy F. Chen, Steven C. H. Hoi

Research Collection School Of Computing and Information Systems

This work is extended from our participation in the Dialogue System Technology Challenge (DSTC7), where we participated in the Audio Visual Scene-aware Dialogue System (AVSD) track. The AVSD track evaluates how dialogue systems understand video scenes and responds to users about the video visual and audio content. We propose a hierarchical attention approach on user queries, video caption, audio and visual features that contribute to improved evaluation results. We also apply a nonlinear feature fusion approach to combine the visual and audio features for better knowledge representation. Our proposed model shows superior performance in terms of both objective evaluation and …

Neural Network Model Of Information Fusion For Coal Storage And Kinetic Energy Of Ball Mill, Bai Yan, He Fang

Journal of System Simulation

Abstract: A dynamic mathematical model of coal pulverizing system was analyzed. Simulation experiments on mill operation process were conducted by PFC3D software platform based on discrete element method. The associated data between different coal quality, coal storage and balls' motion were obtained under certain quantitative optimized operating parameters configuration. Neural network model of information fusion for coal storage and kinetic energy of ball mill was established by using an adaptive combination learning algorithm. Coal storage in mill cylinder was predicted from the energy point of view. The results indicate that there is a close relationship between coal storage, pulverizing efficiency …

Neural Network Inverse Control For The Output Voltage Of Energy Storage Inverter In Micro-Grid, Weiliang Liu, Yongjun Lin, Changliang Liu, Wenying Chen, Liangyu Ma

Journal of System Simulation

Abstract: In order to improve the output voltage waveform quality of energy storage inverter in micro-grid, an inverse control method was proposed based on BP neural network. Mathematical model of the energy storage inverter was established, and the main factors affecting the output voltage were analyzed, and then the expansion inverse model of the system was established based on BP neural network. In order to overcome the local optimum disadvantage in BP training algorithm, gravity algorithm was adopted to optimize the network initial parameters. The neural network inverse model was put in series with its original model to form a …

Soft Sensor Of Particle Size Of Grinding Process Based On Improved Csapso Neural Networks, Zhou Ying, Huimin Zhao, Chen Yang, Wang Long

Journal of System Simulation

Abstract: Aiming at the problems that the particle size can’t be measured online and the offline analysis by lab sample existing in large-time delay, by combining the characteristics of the one stage grinding circuit, the soft sensor model of particle size was proposed by the combination of improved chaotic self-adaptive particle swarm optimization and BP neural network algorithm. Taking advantages of chaotic theory ergodicity and PSO global optimal searching ability, the algorithm above couldadjust the weights of BP network adaptively and avoid falling into the local optimum. As a result of MATLAB simulation, the measurement accuracy of the improved CSAPSO-BP …

Uav Takeoff Decision Based On Neural Network Model Of Takeoff Capability, Yongtao Peng, Yueping Wang, Xiaoting Wang

Journal of System Simulation

Abstract: To enhance the safety in case of engine flameout failure, a new type of UAV takeoff decision based on neural network capacity model was proposed. Two capacity parameters of takeoff safety in case of engine flameout failure were defined, one is the maximum velocity for a safe takeoff and the other is the minimum velocity for a safe shut down. A calculation method based on iterative simulations for those parameters under multiple flight conditions was introduced. Double layer neural networks were used to model the relationship between flight conditions and the capacity parameters, to realize the compressive storage and …

Boiler Combustion Optimization Based On Bayesian Neural Network And Genetic Algorithm, Haiquan Fang, Huifeng Xue, Li Ning, Fei Xi

Journal of System Simulation

Abstract: Neural network and genetic algorithm have been extensively used in boiler combustion optimization problems. But the traditional Back Propagation neural network's generalization ability is poor. The Bayesian regularization can improve the neural network's generalization ability. A boiler combustion multi-objective optimization method combining Bayesian regularization BP neural network and genetic algorithm （Bayes NN-GA）was researched. A number of field test data from a boiler was used to simulate the Bayesian neural network model. The results show that the thermal efficiency and NOx emissions predicted by the Bayesian neural network model show good agreement with the measured, and the optimal results show …

Novel Deep Learning Methods Combined With Static Analysis For Source Code Processing, Duy Quoc Nghi Bui

Dissertations and Theses Collection (Open Access)

It is desirable to combine machine learning and program analysis so that one can leverage the best of both to increase the performance of software analytics. On one side, machine learning can analyze the source code of thousands of well-written software projects that can uncover patterns that partially characterize software that is reliable, easy to read, and easy to maintain. On the other side, the program analysis can be used to define rigorous and unique rules that are only available in programming languages, which enrich the representation of source code and help the machine learning to capture the patterns better. …

Improving Convolutional Neural Network Robustness To Adversarial Images Through Image Filtering, Natalie E. Bogda

Masters Theses

The field of computer vision and deep learning is known for its ability to recognize images with extremely high accuracy. Convolutional neural networks exist that can correctly classify 96\% of 1.2 million images of complex scenes. However, with just a few carefully positioned imperceptible changes to the pixels of an input image, an otherwise accurate network will misclassify this almost identical image with high confidence. These perturbed images are known as \textit{adversarial examples} and expose that convolutional neural networks do not necessarily "see" the world in the way that humans do. This work focuses on increasing the robustness of classifiers …

Secure Mobile Computing By Using Convolutional And Capsule Deep Neural Networks, Rui Ning

Electrical & Computer Engineering Theses & Dissertations

Mobile devices are becoming smarter to satisfy modern user's increasing needs better, which is achieved by equipping divers of sensors and integrating the most cutting-edge Deep Learning (DL) techniques. As a sophisticated system, it is often vulnerable to multiple attacks (side-channel attacks, neural backdoor, etc.). This dissertation proposes solutions to maintain the cyber-hygiene of the DL-Based smartphone system by exploring possible vulnerabilities and developing countermeasures.

First, I actively explore possible vulnerabilities on the DL-Based smartphone system to develop proactive defense mechanisms. I discover a new side-channel attack on smartphones using the unrestricted magnetic sensor data. I demonstrate that attackers can …

Application Of Pso-Bp Algorithm In Hydraulic System Fault Diagnosis, Handong Zhang, Liusong Tao

Journal of System Simulation

Abstract: It is of great significance to monitor, forecast and diagnose hydraulic systems’ fault timely and accurately. First, this paper describes the basic fault model theoretical knowledge of BP neural neystem failure neural network modeling has created and simulated. PSO-BP neural network has been raised, this paper has established PSO optimize model of the BP neural system fault diagnosis. BP network has been created and simulated in Plunger pump hydraulic system failure. The correct results indicate that this mixed PSO-BP algorithm is better than the improved BP algorithm, and can meet the requirements of Hydraulic system fault diagnosis.

Two Power Sliding Mode Neural Network Compensation Control For Space Robot After Target Capturing, Cheng Jing, Chen Li

Journal of System Simulation

Abstract: The impact analyses of space robot capturing a target and stability control problem in the post-impact process were discussed. The dynamic models of space robot system and target were derived by multi-body theory. The impact effect of rigidcouplingmodel was analyzed by applying geometric relationship and principle of momentum conservation. Atwo power sliding mode neural network control scheme was proposed for the combined system after acquiring with uncertain system parameters and external disturbance. The convergence speed of the control system was guaranteed by applyingtwo power sliding mode reaching raw, and the uncertain part was compensated by using neural …

Adaptive Control For Hydraulic Servo Position System With Bounded Input, Jianfei Shi, Shujuan Yi

Journal of System Simulation

Abstract: An adaptive state feedback controller based on neural network fitting was proposed for hydraulic servo position systems containing parameter uncertainties, external disturbance and bounded input problem. Taking the saturation characteristic into account sufficiently, the adaptive state feedback trajectory tracking controller was designed with an adaptive law to real-timely adjust the disturbance parameters and the bounded hyperbolic tangent functions to promise the bounded of the control law. Moreover, the complete stability and performance analysis were presented using Lyapunov theory. Simulation results show the effectiveness of the designed controller for the trajectory tracking in the present of actuators saturation.

Free Space Detection And Trajectory Planning For Autonomous Robot, Zachary Ross Winger

Computer Science and Software Engineering

Autonomous robots need to know what is around them and where it is safe for them to move to. Because having this ability is so important, Dr. Seng and myself have created a model to predict the free space in front of his autonomous robot, Herbie. We then use this prediction to enforce a driving policy to ensure Herbie drives around safely.

Human Facial Emotion Recognition System In A Real-Time, Mobile Setting, Claire Williamson

Honors Theses

The purpose of this project was to implement a human facial emotion recognition system in a real-time, mobile setting. There are many aspects of daily life that can be improved with a system like this, like security, technology and safety.

There were three main design requirements for this project. The first was to get an accuracy rate of 70%, which must remain consistent for people with various distinguishing facial features. The second goal was to have one execution of the system take no longer than half of a second to keep it as close to real time as possible. Lastly, …

Analysis Of Gameplay Strategies In Hearthstone: A Data Science Approach, Connor W. Watson

Theses

In recent years, games have been a popular test bed for AI research, and the presence of Collectible Card Games (CCGs) in that space is still increasing. One such CCG for both competitive/casual play and AI research is Hearthstone, a two-player adversarial game where players seeks to implement one of several gameplay strategies to defeat their opponent and decrease all of their Health points to zero. Although some open source simulators exist, some of their methodologies for simulated agents create opponents with a relatively low skill level. Using evolutionary algorithms, this thesis seeks to evolve agents with a higher skill …

Advancing Performance Of Retail Recommendation Systems, Lisa Leininger, Johnny Gipson, Kito Patterson, Brad Blanchard

SMU Data Science Review

This paper presents two recommendation models, one traditional and one novel, for a retail men's clothing company. J. Hilburn is a custom-fit, menswear clothing company headquartered in Dallas, Texas. J. Hilburn employs stylists across the United States, who engage directly with customers to assist in selecting clothes that fit their size and style. J. Hilburn tasked the authors of this paper to leverage data science techniques to the given data set to provide stylists with more insight into clients’ purchase patterns and increase overall sales. This paper presents two recommendation systems which provide stylists with automatic predictions about possible clothing …

Example-Based Colourization Via Dense Encoding Pyramids, Chufeng Xiao, Chu Han, Zhuming Zhang, Jing Qin, Tien-Tsin Wong, Guoqiang Han, Shengfeng He

Research Collection School Of Computing and Information Systems

We propose a novel deep example-based image colourization method called dense encoding pyramid network. In our study, we define the colourization as a multinomial classification problem. Given a greyscale image and a reference image, the proposed network leverages large-scale data and then predicts colours by analysing the colour distribution of the reference image. We design the network as a pyramid structure in order to exploit the inherent multi-scale, pyramidal hierarchy of colour representations. Between two adjacent levels, we propose a hierarchical decoder–encoder filter to pass the colour distributions from the lower level to higher level in order to take both …

Learning In Feedforward Neural Networks Accelerated By Transfer Entropy, Adrian Moldovan, Angel Caţaron, Rǎzvan Andonie

All Faculty Scholarship for the College of the Sciences

Current neural networks architectures are many times harder to train because of the increasing size and complexity of the used datasets. Our objective is to design more efficient training algorithms utilizing causal relationships inferred from neural networks. The transfer entropy (TE) was initially introduced as an information transfer measure used to quantify the statistical coherence between events (time series). Later, it was related to causality, even if they are not the same. There are only few papers reporting applications of causality or TE in neural networks. Our contribution is an information-theoretical method for analyzing information transfer between the nodes of …

Brian Valdez - Dynamics And Control Of A 3-Dof Manipulator With Deep Learning Feedback, Brian Orlando Valdez

Open Access Theses & Dissertations

With the ever-increasing demands in the space domain and accessibility to low-cost small satellite platforms for educational and scientific projects, efforts are being made in various technology capacities including robotics and artificial intelligence in microgravity. The MIRO Center for Space Exploration and Technology Research (cSETR) prepares the development of their second nanosatellite to launch to space and it is with that opportunity that a 3-DOF robotic arm is in development to be one of the payloads in the nanosatellite. Analyses, hardware implementation, and testing demonstrate a potential positive outcome from including the payload in the nanosatellite and a deep learning …

A Description Of A Humans Knowledge Using Artificial Intelligence, Dj Price

Mahurin Honors College Capstone Experience/Thesis Projects

There currently does not exist a way to easily view the relationships between a collection of written items (e.g. sports articles, diary entries, research papers). In recent years, novel machine learning methods have been developed which are very good at extracting semantic relationships from large numbers of documents. One of them is the (unsupervised) machine learning model Doc2Vec which constructs vectors for documents. The research project detailed in this paper uses this and other already existing algorithms to analyze the relationship between pieces of text. We set forth a broader ambition for this project before discussing the use and need …

Fault Detection And Classification Of A Single Phase Inverter Using Artificial Neural Networks, Ayomikun Samuel Orukotan

All Graduate Theses, Dissertations, and Other Capstone Projects

The detection of switching faults of power converters or the Circuit Under Test (CUT) is real-time important for safe and efficient usage. The CUT is a single-phase inverter. This thesis presents two unique methods that rely on backpropagation principles to solve classification problems with a two-layer network. These mathematical algorithms or proposed networks are able to diagnose single, double, triple, and multiple switching faults over different iterations representing range of frequencies. First, the fault detection and classification problems are formulated as neural network-based classification problems and the neural network design process is clearly described. Then, neural networks are trained over …

Fusion-Net: Integration Of Dimension Reduction And Deep Learning Neural Network For Image Classification, Mohammad Masum, Philippe Laval

Published and Grey Literature from PhD Candidates

Building a deep network using original digital images requires learning many parameters which may reduce the accuracy rates. The images can be compressed by using dimension reduction methods and extracted reduced features can be feeding into a deep network for classification. Hence, in the training phase of the network, the number of parameters will be decreased. Principal Component Analysis is a well-known dimension reduction technique that leverage orthogonal linear transformation of the original data. In this paper, we propose a neural network-based framework, named Fusion-Net, which implements PCA on an image dataset (CIFAR-10) and then a neural network applies on …

A User-Centric And Sentiment Aware Privacy-Disclosure Detection Framework Based On Multi-Input Neural Network, A. K. M. Nuhil Mehdy, Hoda Mehrpouyan

Computer Science Faculty Publications and Presentations

Data and information privacy is a major concern of today’s world. More specifically, users’ digital privacy has become one of the most important issues to deal with, as advancements are being made in information sharing technology. An increasing number of users are sharing information through text messages, emails, and social media without proper awareness of privacy threats and their consequences. One approach to prevent the disclosure of private information is to identify them in a conversation and warn the dispatcher before the conveyance happens between the sender and the receiver. Another way of preventing information (sensitive) loss might be to …

Model-Based Machine Learning To Identify Clinical Relevance In A High-Resolution Simulation Of Sepsis And Trauma, Zachary H. Silberman Md, Robert Chase Cockrell Phd, Gary An Md

Larner College of Medicine Fourth Year Advanced Integration Teaching/Scholarly Projects

Introduction: Sepsis is a devastating, costly, and complicated disease. It represents the summation of varied host immune responses in a clinical and physiological diagnosis. Despite extensive research, there is no current mediator-directed therapy, nor a biomarker panel able to categorize disease severity or reliably predict outcome. Although still distant from direct clinical translation, dynamic computational and mathematical models of acute systemic inflammation and sepsis are being developed. Although computationally intensive to run and calibrate, agent-based models (ABMs) are one type of model well suited for this. New analytical methods to efficiently extract knowledge from ABMs are needed. Specifically, machine-learning …