Physical Sciences and Mathematics Commons^™

Self-Optimizing Feature Generation Via Categorical Hashing Representation And Hierarchical Reinforcement Crossing, Wangyang Ying, Dongjie Wang, Kunpeng Liu, Leilei Sun, Yanjie Fu

Computer Science Faculty Publications and Presentations

Feature generation aims to generate new and meaningful features to create a discriminative representation space. A generated feature is meaningful when the generated feature is from a feature pair with inherent feature interaction. In the real world, experienced data scientists can identify potentially useful feature-feature interactions, and generate meaningful dimensions from an exponentially large search space in an optimal crossing form over an optimal generation path. But, machines have limited human-like abilities. We generalize such learning tasks as self-optimizing feature generation. Self-optimizing feature generation imposes several under-addressed challenges on existing systems: meaningful, robust, and efficient generation. To tackle these challenges, …

From Machine Learning To Deep Learning: A Comprehensive Study Of Alcohol And Drug Use Disorder, Banafsheh Rekabdar, David L. Albright, Haelim Jeong, Sameerah Talafha

Computer Science Faculty Publications and Presentations

This study aims to train and validate machine learning and deep learning models to identify patients with risky alcohol and drug misuse in a Screening, Brief Intervention, and Referral to Treatment (SBIRT) program. An observational cohort of 6978 adults was admitted in the western region of Alabama at three medical facilities between January and December of 2019. Data were cleaned and pre-processed using data imputation techniques and an augmented sampling data method. The primary analysis involved the multi-class classification of alcohol and drug misuse. Our study shows that accurate identification of alcohol and drug use screening instrument scores was best …

A Simpler Machine Learning Model For Acute Kidney Injury Risk Stratification In Hospitalized Patients, Yirui Hu, Kunpeng Liu, Kevin Ho, David Riviello, Jason Brown, Alex R. Chang, Gurmukteshwar Singh, H. Lester Kirchner

Computer Science Faculty Publications and Presentations

Background: Hospitalization-associated acute kidney injury (AKI), affecting one-in-five inpatients, is associated with increased mortality and major adverse cardiac/kidney endpoints. Early AKI risk stratification may enable closer monitoring and prevention. Given the complexity and resource utilization of existing machine learning models, we aimed to develop a simpler prediction model. Methods: Models were trained and validated to predict risk of AKI using electronic health record (EHR) data available at 24 h of inpatient admission. Input variables included demographics, laboratory values, medications, and comorbidities. Missing values were imputed using multiple imputation by chained equations. Results: 26,410 of 209,300 (12.6%) inpatients developed AKI during …

Learning From Machines: Insights In Forest Transpiration Using Machine Learning Methods, Morgan Tholl

Dissertations and Theses

Machine learning has been used as a tool to model transpiration for individual sites, but few models are capable of generalizing to new locations without calibration to site data. Using the global SAPFLUXNET database, 95 tree sap flow data sites were grouped using three clustering strategies: by biome, by tree functional type, and through use of a k-means unsupervised clustering algorithm. Two supervised machine learning algorithms, a random forest algorithm and a neural network algorithm, were used to build machine learning models that predicted transpiration for each cluster. The performance and feature importance in each model were analyzed and compared …

Unpaired Style Transfer Conditional Generative Adversarial Network For Scanned Document Generation, David Jonathan Hawbaker

Dissertations and Theses

Neural networks are a powerful machine learning tool, especially when trained on a large dataset of relevant high-quality data. Generative adversarial networks, image super resolution and most other image manipulation neural networks require a dataset of images and matching target images for training. Collecting and compiling that data can be time consuming and expensive. This work explores an approach for building a dataset of paired document images with a matching scanned version of each document without physical printers or scanners. A dataset of these document image pairs could be used to train a generative adversarial network or image super resolution …

Snerf: Stylized Neural Implicit Representations For 3d Scenes, Thu Nguyen-Phuoc, Feng Liu, Lei Xiao

Computer Science Faculty Publications and Presentations

This paper presents a stylized novel view synthesis method. Applying state-of-the-art stylization methods to novel views frame by frame often causes jittering artifacts due to the lack of cross-view consistency. Therefore, this paper investigates 3D scene stylization that provides a strong inductive bias for consistent novel view synthesis. Specifically, we adopt the emerging neural radiance fields (NeRF) as our choice of 3D scene representation for their capability to render high-quality novel views for a variety of scenes. However, as rendering a novel view from a NeRF requires a large number of samples, training a stylized NeRF requires a large amount …

From Mdp To Alphazero, David Robert Sewell

Dissertations and Theses

In this paper I will explain the AlphaGo family of algorithms starting from first principles and requiring little previous knowledge from the reader. The focus will be upon one of the more recent versions AlphaZero but I hope to explain the core principles that allowed these algorithms to be so successful. I will generally refer to AlphaZero as theses [sic] core set of principles and will make it clear when I am referring to a specific algorithm of the AlphaGo family. AlphaZero in short combines Monte Carlo Tree Search (MCTS) with Deep learning and self-play. We will see how these …

Graphical Models In Reconstructability Analysis And Bayesian Networks, Marcus Harris, Martin Zwick

Systems Science Faculty Publications and Presentations

Reconstructability Analysis (RA) and Bayesian Networks (BN) are both probabilistic graphical modeling methodologies used in machine learning and artificial intelligence. There are RA models that are statistically equivalent to BN models and there are also models unique to RA and models unique to BN. The primary goal of this paper is to unify these two methodologies via a lattice of structures that offers an expanded set of models to represent complex systems more accurately or more simply. The conceptualization of this lattice also offers a framework for additional innovations beyond what is presented here. Specifically, this paper integrates RA and …

Automated Decision Making And Machine Learning: Regulatory Alternatives For Autonomous Settings, Alyssa Heminger

University Honors Theses

Given growing investment capital in research and development, accompanied by extensive literature on the subject by researchers in nearly every domain from civil engineering to legal studies, automated decision-support systems (ADM) are likely to see a place in the foreseeable future. Artificial intelligence (AI), as an automated system, can be defined as a broad range of computerized tasks designed to replicate human neural networks, store and organize large quantities of information, detect patterns, and make predictions with increasing accuracy and reliability. By itself, artificial intelligence is not quite science-fiction tropes (i.e. an uncontrollable existential threat to humanity) yet not without …

View Synthesis Of Dynamic Scenes Based On Deep 3d Mask Volume, Kai-En Lin, Guowei Yang, Lei Xiao, Feng Liu, Ravi Ramamoorthi

Computer Science Faculty Publications and Presentations

Image view synthesis has seen great success in reconstructing photorealistic visuals, thanks to deep learning and various novel representations. The next key step in immersive virtual experiences is view synthesis of dynamic scenes. However, several challenges exist due to the lack of high-quality training datasets, and the additional time dimension for videos of dynamic scenes. To address this issue, we introduce a multi-view video dataset, captured with a custom 10-camera rig in 120FPS. The dataset contains 96 high-quality scenes showing various visual effects and human interactions in outdoor scenes. We develop a new algorithm, Deep 3D Mask Volume, which enables …

Sensitivity Analysis Of An Agent-Based Simulation Model Using Reconstructability Analysis, Andey M. Nunes, Martin Zwick, Wayne Wakeland

Systems Science Faculty Publications and Presentations

Reconstructability analysis, a methodology based on information theory and graph theory, was used to perform a sensitivity analysis of an agent-based model. The NetLogo BehaviorSpace tool was employed to do a full 2^k factorial parameter sweep on Uri Wilensky’s Wealth Distribution NetLogo model, to which a Gini-coefficient convergence condition was added. The analysis identified the most influential predictors (parameters and their interactions) of the Gini coefficient wealth inequality outcome. Implications of this type of analysis for building and testing agent-based simulation models are discussed.

Exploring The Potential Of Sparse Coding For Machine Learning, Sheng Yang Lundquist

Dissertations and Theses

While deep learning has proven to be successful for various tasks in the field of computer vision, there are several limitations of deep-learning models when compared to human performance. Specifically, human vision is largely robust to noise and distortions, whereas deep learning performance tends to be brittle to modifications of test images, including being susceptible to adversarial examples. Additionally, deep-learning methods typically require very large collections of training examples for good performance on a task, whereas humans can learn to perform the same task with a much smaller number of training examples.

In this dissertation, I investigate whether the use …

Leveraging Model Flexibility And Deep Structure: Non-Parametric And Deep Models For Computer Vision Processes With Applications To Deep Model Compression, Anthony D. Rhodes

Dissertations and Theses

My dissertation presents several new algorithms incorporating non-parametric and deep learning approaches for computer vision and related tasks, including object localization, object tracking and model compression. With respect to object localization, I introduce a method to perform active localization by modeling spatial and other relationships between objects in a coherent "visual situation" using a set of probability distributions. I further refine this approach with the Multipole Density Estimation with Importance Clustering (MIC-Situate) algorithm. Next, I formulate active, "situation" object search as a Bayesian optimization problem using Gaussian Processes. Using my Gaussian Process Context Situation Learning (GP-CL) algorithm, I demonstrate improved …

Dictionary Learning For Image Reconstruction Via Numerical Non-Convex Optimization Methods, Lewis M. Hicks

University Honors Theses

This thesis explores image dictionary learning via non-convex (difference of convex, DC) programming and its applications to image reconstruction. First, the image reconstruction problem is detailed and solutions are presented. Each such solution requires an image dictionary to be specified directly or to be learned via non-convex programming. The solutions explored are the DCA (DC algorithm) and the boosted DCA. These various forms of dictionary learning are then compared on the basis of both image reconstruction accuracy and number of iterations required to converge.

An Application Of Deep Learning Models To Automate Food Waste Classification, Alejandro Zachary Espinoza

Dissertations and Theses

Food wastage is a problem that affects all demographics and regions of the world. Each year, approximately one-third of food produced for human consumption is thrown away. In an effort to track and reduce food waste in the commercial sector, some companies utilize third party devices which collect data to analyze individual contributions to the global problem. These devices track the type of food wasted (such as vegetables, fruit, boneless chicken, pasta) along with the weight. Some devices also allow the user to leave the food in a kitchen container while it is weighed, so the container weight must also …

Sensory Relevance Models, Walt Woods

Dissertations and Theses

This dissertation concerns methods for improving the reliability and quality of explanations for decisions based on Neural Networks (NNs). NNs are increasingly part of state-of-the-art solutions for a broad range of fields, including biomedical, logistics, user-recommendation engines, defense, and self-driving vehicles. While NNs form the backbone of these solutions, they are often viewed as "black box" solutions, meaning the only output offered is a final decision, with no insight into how or why that particular decision was made. For high-stakes fields, such as biomedical, where lives are at risk, it is often more important to be able to explain a …

Design And Experimental Evaluation Of Deepmarket: An Edge Computing Marketplace With Distributed Tensorflow Execution Capability, Soyoung Kim

Dissertations and Theses

There is a rise in demand among machine learning researchers for powerful computational resources to train complex machine learning models, e.g., deep learning models. In order to train these models in a reasonable amount of time, the training is often distributed among multiple machines; yet paying for such machines (either through renting them on cloud data centers or building a local infrastructure) is costly. DeepMarket attempts to reduce these costs by creating a marketplace that integrates multiple computational resources over a distributed TensorFlow framework. Instead of requiring users to rent expensive GPU/CPUs from a third-party cloud provider, DeepMarket allows users …

Exploring And Expanding The One-Pixel Attack, Umairullah Khan, Walt Woods, Christof Teuscher

Student Research Symposium

In machine learning research, adversarial examples are normal inputs to a classifier that have been specifically perturbed to cause the model to misclassify the input. These perturbations rarely affect the human readability of an input, even though the model’s output is drastically different. Recent work has demonstrated that image-classifying deep neural networks (DNNs) can be reliably fooled with the modification of a single pixel in the input image, without knowledge of a DNN’s internal parameters. This “one-pixel attack” utilizes an iterative evolutionary optimizer known as differential evolution (DE) to find the most effective pixel to perturb, via the evaluation of …

Spectral Clustering For Electrical Phase Identification Using Advanced Metering Infrastructure Voltage Time Series, Logan Blakely

Dissertations and Theses

The increasing demand for and prevalence of distributed energy resources (DER) such as solar power, electric vehicles, and energy storage, present a unique set of challenges for integration into a legacy power grid, and accurate models of the low-voltage distribution systems are critical for accurate simulations of DER. Accurate labeling of the phase connections for each customer in a utility model is one area of grid topology that is known to have errors and has implications for the safety, efficiency, and hosting capacity of a distribution system. This research presents a methodology for the phase identification of customers solely using …

Knowing Without Knowing: Real-Time Usage Identification Of Computer Systems, Leila Mohammed Hawana

Dissertations and Theses

Contemporary computers attempt to understand a user's actions and preferences in order to make decisions that better serve the user. In pursuit of this goal, computers can make observations that range from simple pattern recognition to listening in on conversations without the device being intentionally active. While these developments are incredibly useful for customization, the inherent security risks involving personal data are not always worth it. This thesis attempts to tackle one issue in this domain, computer usage identification, and presents a solution that identifies high-level usage of a system at any given moment without looking into any personal data. …

Dc-Rts Noise: Observation And Analysis, Benjamin William Hendrickson

Dissertations and Theses

Dark current random telegraph signal (DC-RTS) is a physical phenomenon that effects the performance of solid state image sensors. Identified by meta-stable stochastic switching between two or more dark current levels, DC-RTS is an emerging concern for device scientists and manufacturers as a limiting noise source. Observed and studied in both charge coupled devices (CCDs) and complementary metal-oxide-semiconductor (CMOS) image sensors, the metastable defects inside the device structure that give rise to this switching phenomenon are known to be derived from radiation damage. An examination of the relationship between high energy photon damage and these RTS defects is presented and …

The Silencing Power Of Algorithms: How The Facebook News Feed Algorithm Manipulates Users' Perceptions Of Opinion Climates, Callie Jessica Morgan

University Honors Theses

This extended literature review investigates how the architecture and features of the Facebook Newsfeed algorithm, EdgeRank, can inhibit and facilitate the expression of political opinions. This paper will investigate how Elisabeth Noelle-Neumann's theory on public opinion, Spiral of Silence, can be used to assess the Facebook news feed as a political opinion source that actively shapes users' perceptions of minority and majority opinion climates. The feedback loops created by the algorithm's criteria influences users' decisions to self-censor or express their political opinions with interpersonal connections and unfamiliar connections on the site.

Bounding Box Improvement With Reinforcement Learning, Andrew Lewis Cleland

Dissertations and Theses

In this thesis, I explore a reinforcement learning technique for improving bounding box localizations of objects in images. The model takes as input a bounding box already known to overlap an object and aims to improve the fit of the box through a series of transformations that shift the location of the box by translation, or change its size or aspect ratio. Over the course of these actions, the model adapts to new information extracted from the image. This active localization approach contrasts with existing bounding-box regression methods, which extract information from the image only once. I implement, train, and …

An Exploration Of Linear Classifiers For Unsupervised Spiking Neural Networks With Event-Driven Data, Wesley Chavez

Dissertations and Theses

Object recognition in video has seen giant strides in accuracy improvements in the last few years, a testament to the computational capacity of deep convolutional neural networks. However, this computational capacity of software-based neural networks coincides with high power consumption compared to that of some spiking neural networks (SNNs), up to 300,000 times more energy per synaptic event in IBM's TrueNorth chip, for example. SNNs are also well-suited to exploit the precise timing of event-driven image sensors, which transmit asynchronous "events" only when the luminance of a pixel changes above or below a threshold value. The combination of event-based imagers …

Cox Processes For Counting By Detection, Purnima Rajan, Yongming Ma, Bruno Jedynak

Portland Institute for Computational Science Publications

In this work, doubly stochastic Poisson (Cox) processes and convolutional neural net (CNN) classifiers are used to estimate the number of instances of an object in an image. Poisson processes are well suited to model events that occur randomly in space, such as the location of objects in an image or the enumeration of objects in a scene. The proposed algorithm selects a subset of bounding boxes in the image domain, then queries them for the presence of the object of interest by running a pre-trained CNN classifier. The resulting observations are then aggregated, and a posterior distribution over the …

Gaussian Processes With Context-Supported Priors For Active Object Localization, Bruno Jedynak

Portland Institute for Computational Science Publications

We devise an algorithm using a Bayesian optimization framework in conjunction with contextual visual data for the efficient localization of objects in still images. Recent research has demonstrated substantial progress in object localization and related tasks for computer vision. However, many current state-of-the-art object localization procedures still suffer from inaccuracy and inefficiency, in addition to failing to provide a principled and interpretable system amenable to high-level vision tasks. We address these issues with the current research.

Our method encompasses an active search procedure that uses contextual data to generate initial bounding-box proposals for a target object. We train a convolutional …

Opportunity Identification For New Product Planning: Ontological Semantic Patent Classification, Farshad Madani

Dissertations and Theses

Intelligence tools have been developed and applied widely in many different areas in engineering, business and management. Many commercialized tools for business intelligence are available in the market. However, no practically useful tools for technology intelligence are available at this time, and very little academic research in technology intelligence methods has been conducted to date.

Patent databases are the most important data source for technology intelligence tools, but patents inherently contain unstructured data. Consequently, extracting text data from patent databases, converting that data to meaningful information and generating useful knowledge from this information become complex tasks. These tasks are currently …

A Machine Learning Algorithm For Identifying And Tracking Bacteria In Three Dimensions Using Digital Holographic Microscopy, Manuel Bedrossian, Marwan El-Kholy, Daniel Neamati, Jay Nadeau

Physics Faculty Publications and Presentations

Digital Holographic Microscopy (DHM) is an emerging technique for three-dimensional imaging of microorganisms due to its high throughput and large depth of field relative to traditional microscopy techniques. While it has shown substantial success for use with eukaryotes, it has proven challenging for bacterial imaging because of low contrast and sources of noise intrinsic to the method (e.g. laser speckle). This paper describes a custom written MATLAB routine using machine-learning algorithms to obtain three-dimensional trajectories of live, lab-grown bacteria as they move within an essentially unrestrained environment with more than 90% precision. A fully annotated version of the software used …

Bayesian Optimization For Refining Object Proposals, With An Application To Pedestrian Detection, Anthony D. Rhodes

Student Research Symposium

We devise an algorithm using a Bayesian optimization framework in conjunction with contextual visual data for the efficient localization of objects in still images. Recent research has demonstrated substantial progress in object localization and related tasks for computer vision. However, many current state-of-the-art object localization procedures still suffer from inaccuracy and inefficiency, in addition to failing to successfully leverage contextual data. We address these issues with the current research.

Our method encompasses an active search procedure that uses contextual data to generate initial bounding-box proposals for a target object. We train a convolutional neural network to approximate an offset distance …

The Performance Of Random Prototypes In Hierarchical Models Of Vision, Kendall Lee Stewart

Dissertations and Theses

I investigate properties of HMAX, a computational model of hierarchical processing in the primate visual cortex. High-level cortical neurons have been shown to respond highly to particular natural shapes, such as faces. HMAX models this property with a dictionary of natural shapes, called prototypes, that respond to the presence of those shapes. The resulting set of similarity measurements is an effective descriptor for classifying images. Curiously, prior work has shown that replacing the dictionary of natural shapes with entirely random prototypes has little impact on classification performance. This work explores that phenomenon by studying the performance of random prototypes on …