Other Electrical and Computer Engineering Commons^™

Learning In The Machine: To Share Or Not To Share?, Jordan Ott, Erik Linstead, Nicholas Lahaye, Pierre Baldi

Engineering Faculty Articles and Research

Weight-sharing is one of the pillars behind Convolutional Neural Networks and their successes. However, in physical neural systems such as the brain, weight-sharing is implausible. This discrepancy raises the fundamental question of whether weight-sharing is necessary. If so, to which degree of precision? If not, what are the alternatives? The goal of this study is to investigate these questions, primarily through simulations where the weight-sharing assumption is relaxed. Taking inspiration from neural circuitry, we explore the use of Free Convolutional Networks and neurons with variable connection patterns. Using Free Convolutional Networks, we show that while weight-sharing is a pragmatic optimization …

Dual-Axis Solar Tracker, Bryan Kennedy

All Undergraduate Projects

Renewable energies, and fuels that are not fossil fuel-based, are one of the prolific topics of debate in modern society. With climate change now becoming a primary focus for scientists and innovators of today, one of the areas for the largest amount of potential and growth is that of the capturing and utilization of Solar Energy. This method involves using a mechanical system to track the progression of the sun as it traverses the sky throughout the day. A dual-axis solar tracker such as the one designed and built for this project, can follow the sun both azimuthally and in …

Ensemble Lung Segmentation System Using Deep Neural Networks, Redha A. Ali, Russell C. Hardie, Hussin K. Ragb

Electrical and Computer Engineering Faculty Publications

Lung segmentation is a significant step in developing computer-aided diagnosis (CAD) using Chest Radiographs (CRs). CRs are used for diagnosis of the 2019 novel coronavirus disease (COVID-19), lung cancer, tuberculosis, and pneumonia. Hence, developing a Computer-Aided Detection (CAD) system would provide a second opinion to help radiologists in the reading process, increase objectivity, and reduce the workload. In this paper, we present the implementation of our ensemble deep learning model for lung segmentation. This model is based on the original DeepLabV3+, which is the extended model of DeepLabV3. Our model utilizes various architectures as a backbone of DeepLabV3+, such as …

Deep Siamese Neural Networks For Facial Expression Recognition In The Wild, Wassan Hayale

Electronic Theses and Dissertations

The variation of facial images in the wild conditions due to head pose, face illumination, and occlusion can significantly affect the Facial Expression Recognition (FER) performance. Moreover, between subject variation introduced by age, gender, ethnic backgrounds, and identity can also influence the FER performance. This Ph.D. dissertation presents a novel algorithm for end-to-end facial expression recognition, valence and arousal estimation, and visual object matching based on deep Siamese Neural Networks to handle the extreme variation that exists in a facial dataset. In our main Siamese Neural Networks for facial expression recognition, the first network represents the classification framework, where we …

Corrections To ‘‘Glaciernet: A Deep-Learning Approach For Debris-Covered Glacier Mapping’’, Zhiyuan Xie, Umesh K. Haritashya, Vijayan K. Asari, Brennan W. Young, Michael P. Bishop, Jeffrey S. Kargel

Electrical and Computer Engineering Faculty Publications

In the above article [1], Figure 2 was incorrect. Unfortunately, we mixed the color label of "CONV $\to $ BN $\to $ ReLu" and "Unpooling" in the CNN structure section of Figure 2. The color label of "CONV $\to $ BN $\to $ ReLu" should be orange while the color label of "Unpooling" should be green. Also, the word "Decoder" is misspelled. That same figure with the same error is also used for the graphic abstract. The corrected figure is given here. None of the sections in the figure is modified. The only change is in the color label of …

Ev Charging Behavior Analysis Using Hybrid Intelligence For 5g Smart Grid, Yi Shen, Wei Fang, Feng Ye, Michel Kadoch

Electrical and Computer Engineering Faculty Publications

With the development of the Internet of Things (IoT) and the widespread use of electric vehicles (EV), vehicle-to-grid (V2G) has sparked considerable discussion as an energy-management technology. Due to the inherently high maneuverability of EVs, V2G systems must provide on-demand service for EVs. Therefore, in this work, we propose a hybrid computing architecture based on fog and cloud with applications in 5G-based V2G networks. This architecture allows the bi-directional flow of power and information between schedulable EVs and smart grids (SGs) to improve the quality of service and cost-effectiveness of energy service providers. However, it is very important to select …

Glaciernet: A Deep-Learning Approach For Debris-Covered Glacier Mapping, Zhiyuan Xie, Umesh K. Haritashya, Vijayan K. Asari, Brennan W. Young, Michael P. Bishop, Jeffrey S. Kargel

Electrical and Computer Engineering Faculty Publications

Rising global temperatures over the past decades is directly affecting glacier dynamics. To understand glacier fluctuations and document regional glacier-state trends, glacier-boundary detection is necessary. Debris-covered glacier (DCG) mapping, however, is notoriously difficult using conventional geospatial technology methods. Therefore, in this research for automated DCG mapping, we evaluate the utility of a convolutional neural network (CNN), which is a deep learning feed-forward neural network. The CNN inputs include Landsat satellite images, an Advanced Land Observation Satellite (ALOS) digital elevation model (DEM) and DEM-derived land-surface parameters. Our CNN based deep-learning approach named GlacierNet was designed by appropriately choosing the type, number …

Mitosisnet: End-To-End Mitotic Cell Detection By Multi-Task Learning, Md Zahangir Alom, Theus Aspiras, Tarek M. Taha, Tj Bowen, Vijayan K. Asari

Electrical and Computer Engineering Faculty Publications

Mitotic cell detection is one of the challenging problems in the field of computational pathology. Currently, mitotic cell detection and counting are one of the strongest prognostic markers for breast cancer diagnosis. The clinical visual inspection on histology slides is tedious, error prone, and time consuming for the pathologist. Thus, automatic mitotic cell detection approaches are highly demanded in clinical practice. In this paper, we propose an end-to-end multi-task learning system for mitosis detection from pathological images which is named"MitosisNet". MitosisNet consist of segmentation, detection, and classification models where the segmentation, and detection models are used for mitosis reference region …

Palmprint Gender Classification Using Deep Learning Methods, Minou Khayami

Graduate Theses, Dissertations, and Problem Reports (ETD)

Gender identification is an important technique that can improve the performance of authentication systems by reducing searching space and speeding up the matching process. Several biometric traits have been used to ascertain human gender. Among them, the human palmprint possesses several discriminating features such as principal-lines, wrinkles, ridges, and minutiae features and that offer cues for gender identification. The goal of this work is to develop novel deep-learning techniques to determine gender from palmprint images. PolyU and CASIA palmprint databases with 90,000 and 5502 images respectively were used for training and testing purposes in this research. After ROI extraction and …

Amodal Instance Segmentation And Multi-Object Tracking With Deep Pixel Embedding, Yanfeng Liu

Department of Electrical and Computer Engineering: Dissertations, Theses, and Student Research

This thesis extends upon the representational output of semantic instance segmentation by explicitly including both visible and occluded parts. A fully convolutional network is trained to produce consistent pixel-level embedding across two layers such that, when clustered, the results convey the full spatial extent and depth ordering of each instance. Results demonstrate that the network can accurately estimate complete masks in the presence of occlusion and outperform leading top-down bounding-box approaches.

The model is further extended to produce consistent pixel-level embeddings across two consecutive image frames from a video to simultaneously perform amodal instance segmentation and multi-object tracking. No post-processing …

Bitcoin Price Prediction Using Neural Networks, Vladislav Killiakov

Electrical Engineering

In this project, I will investigate the performance of several major neural network architectures for the task of Bitcoin price prediction. Bitcoin is a cryptocurrency that is recently becoming increasingly more popular, and more widely adopted as a financial instrument. As a result, more efforts have been made in the past several years to model and predict its price. However, to this moment a large portion of work on Bitcoin price modeling was done using statistical or classical machine learning techniques. At the same time, other artificial intelligence based prediction techniques, and specifically neural networks, have not been explored to …

Pixel-Level Deep Multi-Dimensional Embeddings For Homogeneous Multiple Object Tracking, Mateusz Mittek

Department of Electrical and Computer Engineering: Dissertations, Theses, and Student Research

The goal of Multiple Object Tracking (MOT) is to locate multiple objects and keep track of their individual identities and trajectories given a sequence of (video) frames. A popular approach to MOT is tracking by detection consisting of two processing components: detection (identification of objects of interest in individual frames) and data association (connecting data from multiple frames). This work addresses the detection component by introducing a method based on semantic instance segmentation, i.e., assigning labels to all visible pixels such that they are unique among different instances. Modern tracking methods often built around Convolutional Neural Networks (CNNs) and additional, …

The Stability Analysis For Wind Turbines With Doubly Fed Induction Generators, Baohua Dong

Department of Electrical and Computer Engineering: Dissertations, Theses, and Student Research

The quickly increasing, widespread use of wind generation around the world reduces carbon emissions, decreases the effects of global warming, and lowers dependence on fossil fuels. However, the growing penetration of wind power requires more effort to maintain power systems stability.

This dissertation focuses on developing a novel algorithm which dynamically optimizes the proportional-integral (PI) controllers of a doubly fed induction generator (DFIG) driven by a wind turbine to increase the transient performance based on small signal stability analysis.

Firstly, the impact of wind generation is introduced. The stability of power systems with wind generation is described, including the different …

Combining Virtual Reality And Machine Learning For Enhancing The Resiliency Of Transportation Infrastructure In Extreme Events, Supratik Mukhopadhyay, Yimin Zhu, Ravindra Gudishala

Data

Corresponding data set for Tran-SET Project No. 18ITSLSU09. Abstract of the final report is stated below for reference:

"Traffic management models that include route choice form the basis of traffic management systems. High-fidelity models that are based on rapidly evolving contextual conditions can have significant impact on smart and energy efficient transportation. Existing traffic/route choice models are generic and are calibrated on static contextual conditions. These models do not consider dynamic contextual conditions such as the location, failure of certain portions of the road network, the social network structure of population inhabiting the region, route choices made by other drivers, …

Combining Virtual Reality And Machine Learning For Enhancing The Resiliency Of Transportation Infrastructure In Extreme Events, Supratik Mukhopadhyay, Yimin Zhu, Ravindra Gudishala

Publications

Traffic management models that include route choice form the basis of traffic management systems. High-fidelity models that are based on rapidly evolving contextual conditions can have significant impact on smart and energy efficient transportation. Existing traffic/route choice models are generic and are calibrated on static contextual conditions. These models do not consider dynamic contextual conditions such as the location, failure of certain portions of the road network, the social network structure of population inhabiting the region, route choices made by other drivers, extreme conditions, etc. As a result, the model’s predictions are made at an aggregate level and for a …

Utilizing Trajectory Optimization In The Training Of Neural Network Controllers, Nicholas Kimball

Master's Theses

Applying reinforcement learning to control systems enables the use of machine learning to develop elegant and efficient control laws. Coupled with the representational power of neural networks, reinforcement learning algorithms can learn complex policies that can be difficult to emulate using traditional control system design approaches. In this thesis, three different model-free reinforcement learning algorithms, including Monte Carlo Control, REINFORCE with baseline, and Guided Policy Search are compared in simulated, continuous action-space environments. The results show that the Guided Policy Search algorithm is able to learn a desired control policy much faster than the other algorithms. In the inverted pendulum …

Cooperative Learning For The Consensus Of Multi-Agent Systems, Qishuai Liu

Department of Electrical and Computer Engineering: Dissertations, Theses, and Student Research

Due to a lot of attention for the multi-agent system in recent years, the consensus algorithm gained immense popularity for building fault-tolerant systems in system and control theory. Generally, the consensus algorithm drives the swarm of agents to work as a coherent group that can reach an agreement regarding a certain quantity of interest, which depends on the state of all agents themselves. The most common consensus algorithm is the average consensus, the final consensus value of which is equal to the average of the initial values. If we want the agents to find the best area of the particular …

Embedded System Design Of Robot Control Architectures For Unmanned Agricultural Ground Vehicles, Ryan Humphrey

Department of Electrical and Computer Engineering: Dissertations, Theses, and Student Research

Engineering technology has matured to the extent where accompanying methods for unmanned field management is now becoming a technologically achievable and economically viable solution to agricultural tasks that have been traditionally performed by humans or human operated machines. Additionally, the rapidly increasing world population and the daunting burden it places on farmers in regards to the food production and crop yield demands, only makes such advancements in the agriculture industry all the more imperative. Consequently, the sector is beginning to observe a noticeable shift, where there exist a number of scalable infrastructural changes that are in the process of slowly …

Aerial Base Station Deployment For Post-Disaster Public Safety Applications, Jose Antonio Matamoros Vargas

Department of Electrical and Computer Engineering: Dissertations, Theses, and Student Research

Earthquakes and floods are constant threats to most of the countries in the world. After such catastrophes, a rapid response is needed, which includes communications not only for first responders but also for local civilians. Even though there are technologies and specialized personnel for rapid deployment, it is common that external factors will hinder the arrival of help while communication requirements are urgently required. Such communication technologies would aid tasks regarding organization and information dissemination from authorities to the civilians and vice-versa. This necessity is due to protocols and applications to allocate the number of emergency resources per location and …

A State-Of-The-Art Survey On Deep Learning Theory And Architectures, Md Zahangir Alom, Tarek M. Taha, Christopher Yakopcic, Stefan Westberg, Paheding Sidike, Mst Shamima Nasrin, Mahmudul Hasan, Brian C. Van Essen, Abdul A. S. Awwal, Vijayan K. Asari

Electrical and Computer Engineering Faculty Publications

In recent years, deep learning has garnered tremendous success in a variety of application domains. This new field of machine learning has been growing rapidly and has been applied to most traditional application domains, as well as some new areas that present more opportunities. Different methods have been proposed based on different categories of learning, including supervised, semi-supervised, and un-supervised learning. Experimental results show state-of-the-art performance using deep learning when compared to traditional machine learning approaches in the fields of image processing, computer vision, speech recognition, machine translation, art, medical imaging, medical information processing, robotics and control, bioinformatics, natural language …

Applications Of Supervised Machine Learning In Autism Spectrum Disorder Research: A Review, Kayleigh K. Hyde, Marlena N. Novack, Nicholas Lahaye, Chelsea Parlett-Pelleriti, Raymond Anden, Dennis R. Dixon, Erik Linstead

Engineering Faculty Articles and Research

Autism spectrum disorder (ASD) research has yet to leverage "big data" on the same scale as other fields; however, advancements in easy, affordable data collection and analysis may soon make this a reality. Indeed, there has been a notable increase in research literature evaluating the effectiveness of machine learning for diagnosing ASD, exploring its genetic underpinnings, and designing effective interventions. This paper provides a comprehensive review of 45 papers utilizing supervised machine learning in ASD, including algorithms for classification and text analysis. The goal of the paper is to identify and describe supervised machine learning trends in ASD literature as …

Active Recall Networks For Multiperspectivity Learning Through Shared Latent Space Optimization, Theus Aspiras, Ruixu Liu, Vijayan K. Asari

Electrical and Computer Engineering Faculty Publications

Given that there are numerous amounts of unlabeled data available for usage in training neural networks, it is desirable to implement a neural network architecture and training paradigm to maximize the ability of the latent space representation. Through multiple perspectives of the latent space using adversarial learning and autoencoding, data requirements can be reduced, which improves learning ability across domains. The entire goal of the proposed work is not to train exhaustively, but to train with multiperspectivity. We propose a new neural network architecture called Active Recall Network (ARN) for learning with less labels by optimizing the latent space. This …

Deep Temporal Convolutional Networks For Short-Term Traffic Flow Forecasting, Wentian Zhao, Yanyun Gao, Tingxiang Ji, Xili Wan, Feng Ye, Guangwei Bai

Electrical and Computer Engineering Faculty Publications

To reduce the increasingly congestion in cities, it is essential for intelligent transportation system (ITS) to accurately forecast the short-term traffic flow to identify the potential congestion sites. In recent years, the emerging deep learning method has been introduced to design traffic flow predictors, such as recurrent neural network (RNN) and long short-term memory (LSTM), which has demonstrated its promising results. In this paper, different from existing work, we study the temporal convolutional network (TCN) and propose a deep learning framework based on TCN model for short-term city-wide traffic forecast to accurately capture the temporal and spatial evolution of traffic …

Effects Of Correlation Of Channel Gains On The Secrecy Capacity In The Gaussian Wiretap Channel, Abhishek Lokur

Department of Electrical and Computer Engineering: Dissertations, Theses, and Student Research

Secrecy capacity is one of the most important characteristic of a wireless communication channel. Therefore, the study of this characteristic wherein the system has correlated channel gains and study them for different line-of-sight (LOS) propagation scenarios is of ultimate importance.

The primary objective of this thesis from the mathematical side is to determine the secrecy capacity (SC) for correlated channel gains for the main and eavesdropper channels in a Gaussian Wiretap channel as a function from main parameters (μ, Σ, ρ). f(h₁, h₂) is the joint distribution of the two channel gains at channel use (h …

Recurrent Residual U-Net For Medical Image Segmentation, Md Zahangir Alom, Christopher Yakopcic, Mahmudul Hasan, Tarek M. Taha, Vijayan K. Asari

Electrical and Computer Engineering Faculty Publications

Deep learning (DL)-based semantic segmentation methods have been providing state-of-the-art performance in the past few years. More specifically, these techniques have been successfully applied in medical image classification, segmentation, and detection tasks. One DL technique, U-Net, has become one of the most popular for these applications. We propose a recurrent U-Net model and a recurrent residual U-Net model, which are named RU-Net and R2U-Net, respectively. The proposed models utilize the power of U-Net, residual networks, and recurrent convolutional neural networks. There are several advantages to using these proposed architectures for segmentation tasks. First, a residual unit helps when training deep …

A Survey Of Techniques For Mobile Service Encrypted Traffic Classification Using Deep Learning, Pan Wang, Xuejiao Chen, Feng Ye, Zhixin Sun

Electrical and Computer Engineering Faculty Publications

The rapid adoption of mobile devices has dramatically changed the access to various net- working services and led to the explosion of mobile service traffic. Mobile service traffic classification has been a crucial task that attracts strong interest in mobile network management and security as well as machine learning communities for past decades. However, with more and more adoptions of encryption over mobile services, it brings a lot of challenges about mobile traffic classification. Although classical machine learning approaches can solve many issues that port and payload-based methods cannot solve, it still has some limitations, such as time-consuming, costly handcrafted …

On Learning And Visualizing Lexicographic Preference Trees, Ahmed S. Moussa

UNF Graduate Theses and Dissertations

Preferences are very important in research fields such as decision making, recommendersystemsandmarketing. The focus of this thesis is on preferences over combinatorial domains, which are domains of objects configured with categorical attributes. For example, the domain of cars includes car objects that are constructed withvaluesforattributes, such as ‘make’, ‘year’, ‘model’, ‘color’, ‘body type’ and ‘transmission’.Different values can instantiate an attribute. For instance, values for attribute ‘make’canbeHonda, Toyota, Tesla or BMW, and attribute ‘transmission’ can haveautomaticormanual. To this end,thisthesis studiesproblemsonpreference visualization and learning for lexicographic preference trees, graphical preference models that often are compact over complex domains of objects built of …

Design And Analysis Of Binary Driven Coherent M-Ary Qam Transmitter For Next Generation Optical Networks, Naji Albakay

Department of Electrical and Computer Engineering: Dissertations, Theses, and Student Research

This work presents a design for a binary driven optical square M-ary quadrature amplitude modulation (QAM) transmitter for high speed optical networks. The transmitter applies tandem quadrature phase shift keying (QPSK) modulators to eliminate the need for linear broadband amplifiers and high-resolution digital to analog converters (DACs), which are both required by conventional transmitters. The transmitter design could be scaled to any order of square M-ary QAM by simply adding more QPSK modulators in tandem. It also provides a Gray coded symbol constellation, insuring the lowest bit error rate possible during symbol recovery. We also provide the design for the …

Logging, Visualization, And Analysis Of Network And Power Data Of Iot Devices, Neal Huynh Nguyen

Master's Theses

There are approximately 23.14 billion IoT(Internet of Things) devices currently in use worldwide. This number is projected to grow to over 75 billion by 2025. Despite their ubiquity little is known about the security and privacy implications of IoT devices. Several large-scale attacks against IoT devices have already been recorded.

To help address this knowledge gap, we have collected a year’s worth of network traffic and power data from 16 common IoT devices. From this data, we show that we can identify different smart speakers, like the Echo Dot, from analyzing one minute of power data on a shared power …

Asic Design To Support Low Power High Voltage Power Supply For Radiation Monitoring Applications, Daniel Rogge

Department of Electrical and Computer Engineering: Dissertations, Theses, and Student Research

A low power high voltage power supply is designed for use in a long duration radiation monitoring system. The supply employs a flexible pulse frequency modulation switching controller implemented in a 0.35 $\mu$m CMOS technology. The controller drives and regulates a flyback transformer driven 12-stage Cockroft-Walton voltage multiplier chain. The chain provides bias for the dynodes of a photomultiplier tube. The supply voltage is selectable via a 12-bit on-chip digital to analog converter. The system is designed for low power operation and immunity to supply voltage variation as the application is battery-powered.

Advisors: Sina Balkir and Michael W. Hoffman