Artificial Intelligence and Robotics Commons^™

Mathematical Models Yield Insights Into Cnns: Applications In Natural Image Restoration And Population Genetics, Ryan Cecil

Electronic Theses and Dissertations

Due to a rise in computational power, machine learning (ML) methods have become the state-of-the-art in a variety of fields. Known to be black-box approaches, however, these methods are oftentimes not well understood. In this work, we utilize our understanding of model-based approaches to derive insights into Convolutional Neural Networks (CNNs). In the field of Natural Image Restoration, we focus on the image denoising problem. Recent work have demonstrated the potential of mathematically motivated CNN architectures that learn both `geometric' and nonlinear higher order features and corresponding regularizers. We extend this work by showing that not only can geometric features …

Classification Models For 2,4-D Formulations In Damaged Enlist Crops Through The Application Of Ftir Spectroscopy And Machine Learning Algorithms, Benjamin Blackburn

Theses and Dissertations

With new 2,4-Dichlorophenoxyacetic acid (2,4-D) tolerant crops, increases in off-target movement events are expected. New formulations may mitigate these events, but standard lab techniques are ineffective in identifying these 2,4-D formulations. Using Fourier-transform infrared spectroscopy and machine learning algorithms, research was conducted to classify 2,4-D formulations in treated herbicide-tolerant soybeans and cotton and observe the influence of leaf treatment status and collection timing on classification accuracy. Pooled Classification models using k-nearest neighbor classified 2,4-D formulations with over 65% accuracy in cotton and soybean. Tissue collected 14 DAT and 21 DAT for cotton and soybean respectively produced higher accuracies than the …

3d Vision With Transformers: A Survey, Jean Lahoud, Jiale Cao, Fahad Shahbaz Khan, Hisham Cholakkal, Rao Anwer, Salman Khan, Ming-Hsuan Yang

Computer Vision Faculty Publications

The success of the transformer architecture in natural language processing has recently triggered attention in the computer vision field. The transformer has been used as a replacement for the widely used convolution operators, due to its ability to learn long-range dependencies. This replacement was proven to be successful in numerous tasks, in which several state-of-the-art methods rely on transformers for better learning. In computer vision, the 3D field has also witnessed an increase in employing the transformer for 3D convolution neural networks and multi-layer perceptron networks. Although a number of surveys have focused on transformers in vision in general, 3D …

Evaluating The Variable Stride Algorithm In The Identification Of Diabetic Retinopathy, Ying Zheng, Brian Danaher, Matthew Brown

Beyond: Undergraduate Research Journal

An experiment was performed to investigate a modified pooling method for use in convolutional neural networks for image recognition. This algorithm–Variable Stride–allows the user to segment an image and change the amount of subsampling in each region. This control allows for the user to maintain a higher amount of data retention in more important regions of the image, while more aggressively subsampling the less important regions to increase training speed. Three Variable Stride methods were compared to the preexisting pooling algorithms, Maximum Pool and Average Pool, in three different network configurations tasked with classifying Diabetic Retinopathy images between its early …

A Multi-Dimensional Matrix Pencil-Based Channel Prediction Method For Massive Mimo With Mobility, Weidong Li, Haifan Yin, Ziao Qin, Yandi Cao, Mérouane Debbah

Machine Learning Faculty Publications

This paper addresses the mobility problem in massive multiple-input multiple-output systems, which leads to significant performance losses in the practical deployment of the fifth generation mobile communication networks. We propose a novel channel prediction method based on multi-dimensional matrix pencil (MDMP), which estimates the path parameters by exploiting the angular-frequency-domain and angular-timedomain structures of the wideband channel. The MDMP method also entails a novel path pairing scheme to pair the delay and Doppler, based on the super-resolution property of the angle estimation. Our method is able to deal with the realistic constraint of time-varying path delays introduced by user movements, …

Artificial Intelligence In The Radiomic Analysis Of Glioblastomas: A Review, Taxonomy, And Perspective, Ming Zhu, Sijia Li, Yu Kuang, Virginia B. Hill, Amy B. Heimberger, Lijie Zhai, Shenjie Zhai

Electrical & Computer Engineering Faculty Research

Radiological imaging techniques, including magnetic resonance imaging (MRI) and positron emission tomography (PET), are the standard-of-care non-invasive diagnostic approaches widely applied in neuro-oncology. Unfortunately, accurate interpretation of radiological imaging data is constantly challenged by the indistinguishable radiological image features shared by different pathological changes associated with tumor progression and/or various therapeutic interventions. In recent years, machine learning (ML)-based artificial intelligence (AI) technology has been widely applied in medical image processing and bioinformatics due to its advantages in implicit image feature extraction and integrative data analysis. Despite its recent rapid development, ML technology still faces many hurdles for its broader applications …

Data Collection And Machine Learning Methods For Automated Pedestrian Facility Detection And Mensuration, Joseph Bailey Luttrell Iv

Dissertations

Large-scale collection of pedestrian facility (crosswalks, sidewalks, etc.) presence data is vital to the success of efforts to improve pedestrian facility management, safety analysis, and road network planning. However, this kind of data is typically not available on a large scale due to the high labor and time costs that are the result of relying on manual data collection methods. Therefore, methods for automating this process using techniques such as machine learning are currently being explored by researchers. In our work, we mainly focus on machine learning methods for the detection of crosswalks and sidewalks from both aerial and street-view …

Data-Driven Research On Engineering Design Thinking And Behaviors In Computer-Aided Systems Design: Analysis, Modeling, And Prediction, Molla Hafizur Rahman

Graduate Theses and Dissertations

Research on design thinking and design decision-making is vital for discovering and utilizing beneficial design patterns, strategies, and heuristics of human designers in solving engineering design problems. It is also essential for the development of new algorithms embedded with human intelligence and can facilitate human-computer interactions. However, modeling design thinking is challenging because it takes place in the designer’s mind, which is intricate, implicit, and tacit. For an in-depth understanding of design thinking, fine-grained design behavioral data are important because they are the critical link in studying the relationship between design thinking, design decisions, design actions, and design performance. Therefore, …

Evaluation Of Generative Models For Predicting Microstructure Geometries In Laser Powder Bed Fusion Additive Manufacturing, Andy Ramlatchan

Computer Science Theses & Dissertations

In-situ process monitoring for metals additive manufacturing is paramount to the successful build of an object for application in extreme or high stress environments. In selective laser melting additive manufacturing, the process by which a laser melts metal powder during the build will dictate the internal microstructure of that object once the metal cools and solidifies. The difficulty lies in that obtaining enough variety of data to quantify the internal microstructures for the evaluation of its physical properties is problematic, as the laser passes at high speeds over powder grains at a micrometer scale. Imaging the process in-situ is complex …

Emotion Detection Using An Ensemble Model Trained With Physiological Signals And Inferred Arousal-Valence States, Matthew Nathanael Gray

Electrical & Computer Engineering Theses & Dissertations

Affective computing is an exciting and transformative field that is gaining in popularity among psychologists, statisticians, and computer scientists. The ability of a machine to infer human emotion and mood, i.e. affective states, has the potential to greatly improve human-machine interaction in our increasingly digital world. In this work, an ensemble model methodology for detecting human emotions across multiple subjects is outlined. The Continuously Annotated Signals of Emotion (CASE) dataset, which is a dataset of physiological signals labeled with discrete emotions from video stimuli as well as subject-reported continuous emotions, arousal and valence, from the circumplex model, is used for …

Scheduling, Complexity, And Solution Methods For Space Robot On-Orbit Servicing, Susan E. Sorenson

Graduate Theses and Dissertations

This research proposes problems, models, and solutions for the scheduling of space robot on-orbit servicing. We present the Multi-Orbit Routing and Scheduling of Refuellable On-Orbit Servicing Space Robots problem which considers on-orbit servicing across multiple orbits with moving tasks and moving refuelling depots. We formulate a mixed integer linear program model to optimize the routing and scheduling of robot servicers to accomplish on-orbit servicing tasks. We develop and demonstrate flexible algorithms for the creation of the model parameters and associated data sets. Our first algorithm creates the network arcs using orbital mechanics. We have also created a novel way to …

Deep Active Genetic Learning With Evidential Uncertainty For Agriculture Crops And Lake Water Quality Assessment, Oguz M. Aranay

Legacy Theses & Dissertations (2009 - 2024)

Despite significant advancements in the field of machine learning, there are two issues that still require further exploration. First, how to learn from a small dataset; and second, how to select appropriate features from the data. Although there exist many techniques to address these issues, choosing a combination of the techniques from these two groups is challenging, and worth investigating. To address these concerns, this thesis presents a learning framework that is based on a deep learning model utilizing active learning (with evidential uncertainty as a basis for acquisition function) for the first issue and a genetic algorithm for the …

Protein-Protein Interaction Prediction From Language Of Biological Coding, Nayan Howladar

LSU New Orleans Theses and Dissertations

Protein-protein interactions in a cell are essential to the characterization and performance of various fundamental biological processes. Due to the tedious, resource-expensive, and time-consuming experimental processes, computational techniques to solve protein pair interaction difficulties have emerged as an active research area in bioinformatics. This research seeks to develop an innovative machine learning-based technique that predicts the interaction of a protein pair based on carefully selected input features and exploits information-rich evolutionary information. We developed a protein-protein interaction predictor, PPILS, that leverages the evolutionary knowledge from the protein language model. We examined several distinct neural network architectures: CNN+LSTM, Transformer, Encoder-Decoder, and …

Parallel Algorithms For Scalable Graph Mining: Applications On Big Data And Machine Learning, Naw Safrin Sattar

LSU New Orleans Theses and Dissertations

Parallel computing plays a crucial role in processing large-scale graph data. Complex network analysis is an exciting area of research for many applications in different scientific domains e.g., sociology, biology, online media, recommendation systems and many more. Graph mining is an area of interest with diverse problems from different domains of our daily life. Due to the advancement of data and computing technologies, graph data is growing at an enormous rate, for example, the number of links in social networks is growing every millisecond. Machine/Deep learning plays a significant role for technological accomplishments to work with big data in modern …

Ocean Wave Prediction And Characterization For Intelligent Maritime Transportation, Pujan Pokhrel

LSU New Orleans Theses and Dissertations

The national Earth System Prediction (ESPC) initiative aims to develop the predictions
for the next generation predictions of atmosphere, ocean, and sea-ice interactions in the scale of days to decades. This dissertation seeks to demonstrate the methods we can use to improve the ESPC models, especially the ocean prediction model. In the application side of the weather forecasts, this dissertation explores imitation learning with constraints to solve combinatorial optimization problems, focusing on the weather routing of surface vessels. Prediction of ocean waves is essential for various purposes, including vessel routing, ocean energy harvesting, agriculture, etc. Since the machine learning approaches …

Directed Acyclic Graph-Based Neural Networks For Tunable Low-Power Computer Vision, Abhinav Goel, Caleb Tung, Nick Eliopoulos, Xiao Hu, George K. Thiruvathukal, James C. Davis, Yung-Hisang Lu

Computer Science: Faculty Publications and Other Works

Processing visual data on mobile devices has many applications, e.g., emergency response and tracking. State-of-the-art computer vision techniques rely on large Deep Neural Networks (DNNs) that are usually too power-hungry to be deployed on resource-constrained edge devices. Many techniques improve DNN efficiency of DNNs by compromising accuracy. However, the accuracy and efficiency of these techniques cannot be adapted for diverse edge applications with different hardware constraints and accuracy requirements. This paper demonstrates that a recent, efficient tree-based DNN architecture, called the hierarchical DNN, can be converted into a Directed Acyclic Graph-based (DAG) architecture to provide tunable accuracy-efficiency tradeoff options. We …

Deep Learning Applications In Industrial And Systems Engineering, Winthrop Harvey

Graduate Theses and Dissertations

Deep learning - the use of large neural networks to perform machine learning - has transformed the world. As the capabilities of deep models continue to grow, deep learning is becoming an increasingly valuable and practical tool for industrial engineering. With its wide applicability, deep learning can be turned to many industrial engineering tasks, including optimization, heuristic search, and functional approximation. In this dissertation, the major concepts and paradigms of deep learning are reviewed, and three industrial engineering projects applying these methods are described. The first applies a deep convolutional network to the task of absolute aerial geolocalization - the …

Deep Learning Edge Detection In Image Inpainting, Zheng Zheng

Electronic Theses, Projects, and Dissertations

In recent years, deep learning has grown rapidly, and it has been creatively implemented for various applications. In 2019, deep learning based EdgeConnect image inpainting algorithm came out and occupied a place in the image inpainting field. Unlike traditional image inpainting methods which mainly read and use the color information of the remaining part of the image to fill the missing regions of the image, EdgeConnect uses the innovative edge-first and color-next approach. It uses an edge detector to generate an edge map of an image with missing regions, then the missing edges are completed by an edge model, finally …

Is Artificial Intelligence A Double-Edged Sword? Insights From Three Essays On Its Impacts, Ankur Arora

Graduate Theses and Dissertations

"If we do it right, we might be able to evolve a form of work that taps into our uniquely human capabilities and restores our humanity. The ultimate paradox is that this technology may become a powerful catalyst that we need to reclaim our humanity." - John Hagel

Artificial intelligence (AI) is viewed as a disruptive technology that some executives believe will take over a lot of jobs. However, others believe that AI will bolster growth, improve business processes, and create new business opportunities. This dissertation focuses on the tension arising from such contrasting expected impacts of AI. Extant research …

Distributed Learning With Automated Stepsizes, Benjamin Liggett

All Theses

Stepsizes for optimization problems play a crucial role in algorithm convergence, where the stepsize must undergo tedious manual tuning to obtain near-optimal convergence. Recently, an adaptive method for automating stepsizes was proposed for centralized optimization. However, this method is not directly applicable to decentralized optimization because it allows for heterogeneous agent stepsizes. Furthermore, directly using consensus between agent stepsizes to mitigate stepsize heterogeneity can decrease performance and even lead to divergence.

This thesis proposes an algorithm to remedy the tedious manual tuning of stepsizes in decentralized optimization. Our proposed algorithm automates the stepsize and uses dynamic consensus between agents’ stepsizes …