ผลกระทบของการจำแนกประเภทแบบสองลาเบลต่อประสิทธิภาพในเรียนรู้ของโครงข่ายประสาทเทียมแบบป้อนไปข้างหน้า, ภานุพงศ์ จิรเศรษฐสิริ

Chulalongkorn University Theses and Dissertations (Chula ETD)

การศึกษานี้พิจารณาผลกระทบของตัวแบบในการจำแนกประเภทแบบสองลาเบลต่อประสิทธิภาพของโครงข่ายประสาทเทียมแบบป้อนไปข้างหน้า โดยเปรียบเทียบกับตัวแบบในการจำแนกประเภทแบบหนึ่งลาเบล การวิจัยนี้มุ่งเน้นไปที่ตัวชี้วัดประสิทธิภาพที่สำคัญ เช่น ความเร็วในการลู่เข้าของตัวแบบ การลดค่าการสูญเสีย และความเสถียรของตัวแบบผ่านการทดสอบในกรณีศึกษาที่มีความซับซ้อนต่างกัน 3 กรณี ผลการวิจัยแสดงให้เห็นว่าการจำแนกประเภทแบบสองลาเบลสามารถทำให้ตัวแบบลู่เข้ารวดเร็วขึ้น มีค่าการสูญเสียต่ำกว่า และมีความเสถียรมากกว่าเมื่อเปรียบเทียบกับตัวแบบในการจำแนกประเภทแบบหนึ่งลาเบล โดยเฉพาะในกรณีของชุดข้อมูลที่มีความซับซ้อนน้อยกว่า อย่างไรก็ตามเมื่อความซับซ้อนของข้อมูลเพิ่มขึ้น ความแตกต่างในด้านประสิทธิภาพระหว่างสองตัวแบบจะเริ่มลดลง แต่การจำแนกประเภทแบบสองลาเบลยังคงมีข้อได้เปรียบเล็กน้อยในด้านการลดค่าการสูญเสีย ความสามารถในการปรับตัว และความสามารถในการจัดการกับชุดข้อมูลที่หลากหลาย ผลการศึกษานี้ชี้ให้เห็นว่าการจำแนกประเภทแบบสองลาเบลสามารถเพิ่มประสิทธิภาพของตัวแบบการเรียนรู้เชิงลึกได้อย่างมีนัยสำคัญ ทำให้สามารถแก้ไขปัญหาที่เกี่ยวข้องกับตัวแปรตามหลายตัวได้อย่างมีประสิทธิภาพ ผลลัพธ์จากการศึกษานี้มีความสำคัญอย่างยิ่งในด้านการวิเคราะห์เชิงพยากรณ์และการจัดประเภทข้อมูลขนาดใหญ่ ซึ่งความเร็วในการประมวลผลและความเสถียรของตัวแบบเป็นปัจจัยสำคัญ

การปรับปรุงแบบจําลองการเรียนรู้เชิงลึกสําหรับการทํานายผลตอบแทนในตลาดหุ้นไทยด้วยข้อมูลทางการเงินและบทความข่าว, ภัคพล ตันติวิชช์

Chulalongkorn University Theses and Dissertations (Chula ETD)

ในปัจจุบันมีการนำข้อมูลหลากหลายประเภทมาใช้ในการทำนายผลตอนแทนในตลาดหุ้น ได้แก่ ข้อมูลราคาหุ้นในอดีต ตัวชี้วัดทางเทคนิค ข้อมูลทางการเงินในอดีต และบทความข่าว อย่างไรก็ตามงานวิจัยส่วนใหญ่มุ่งเน้นการศึกษาอยู่ในตลาดหุ้นต่างประเทศ และมักจะใช้เพียงข้อมูลราคาหุ้นในอดีตและตัวชี้วัดทางเทคนิคเท่านั้น ข้อมูลทั้ง 4 ประเภทข้างต้นยังไม่เคยถูกใช้ร่วมกัน ในการศึกษาครั้งนี้มีวัตถุประสงค์เพื่อนำเสนอวิธีการทำนายผลตอบแทนโดยมุ่งเน้นที่ตลาดหุ้นไทย ด้วยแบบจำลองการเรียนรู้เชิงลึกโดยใช้ข้อมูลนำเข้าทั้ง 4 ประเภท ซึ่งแบบจำลองที่ใช้คือแบบจำลองการเรียนรู้เชิงลึก Temporal Convolutional Networks (TCN) และนำเสนอการปรับปรุงประสิทธิภาพของแบบจำลองทั้งในด้านการทำนายผลตอบแทนและในด้านการสร้างผลตอบแทนจากการจำลองการซื้อขายด้วยกลไกความสนใจ (Attention Mechanism) ในชื่อแบบจำลองการเรียนรู้เชิงลึก Temporal Convolutional Attention Networks (TCAN) และการใช้ข้อมูลย้อนหลัง ผลการศึกษาแสดงให้เห็นถึงความสามารถของแบบจำลอง TCAN ที่ใช้ข้อมูลเพียง 3 ประเภท คือ ข้อมูลราคาหุ้นในอดีต ตัวชี้วัดทางเทคนิค และบทความข่าวที่ถูกนำมาคำนวณเป็นคะแนนความรู้สึก ด้วยข้อมูลย้อนหลังจำนวน 10 วัน เป็นแบบจำลองที่มีประสิทธิภาพในการทำนายและสร้างผลตอบแทนในตลาดหุ้นไทย

Formulating An Efficient Statistical Test Using The Goodness Of Fit Approach With Applications To Real-Life Data, S. A. Qaid, S. E. Abo Youssef Prof., Mahmoud Mansour

Basic Science Engineering

Statistical tests are very important for researchers to make decisions. In particular, when the tests are non-parametric, they are of greater importance because they can be applied to a wide range of data sets regardless of knowing the distribution of these data. Researchers are therefore racing to obtain efficient tests for making good decisions based on the results of these tests. In this study, NBU (2)L was used based on the goodness of fit approach to present an efficient statistical test. The efficiency of the proposed test was computed, and the results were compared to those of other tests. Critical …

The Distribution Of The Significance Level, Paul O. Monnu

College of Graduate Studies: Theses & Dissertations

Reporting the p-value is customary when conducting a test of hypothesis or significance. The likelihood of getting a fictitious second sample and presuming the null hypothesis is correct is the p-value. The significance level is a statistic that interests us to investigate. Being a statistic, it has a distribution. For the F-test in a one-way ANOVA and the t-tests for population means, we define the significance level, its observed value, and the observed significance level. It is possible to derive the significance level distribution. The t-test and the F-test are not without controversy. Specifically, we demonstrate that as sample size …

Investigating Flash Flood Occurrence Using Negative Binomial Models In Maryland, United States Of America, Zainab O. Akinsemoyin

College of Graduate Studies: Theses & Dissertations

Globally, as extreme weather patterns intensify, flash floods have emerged as one of the most destructive and immediate environmental threats. In Maryland, flash floods are particularly concerning due to its diverse topography and increasing urban development, which exacerbates runoff and overwhelms drainage systems. The state has experienced significant flash flood events, highlighting the need for effective models to manage risks and inform mitigation strategies. While regression models such as the Negative Binomial (NB) and Zero-Inflated Negative Binomial (ZINB) are commonly used for count data analysis, their application to flash flood modeling in the USA, including regions like Maryland, remains limited …

Self-Exciting Point Processes In Real Estate, Ian Fraser

Theses and Dissertations (Comprehensive)

This thesis introduces a novel approach to analyzing residential property sales through the lens of stochastic processes by employing point processes. Herein, property sales are treated as point patterns, using self-exciting point process models and a variety of statistical tools to uncover underlying patterns in the data. Key findings include the identification and explanation of clustering in both space and time, and the efficacy of a temporal Hawkes process with a sinusoidal background in predicting home sale occurrences. The temporal analysis starts by employing the state of art techniques for time series data like regression, autoregressive, and autoregressive integrated moving …

Reinforcement Learning: Applying Low Discrepancy Action Selection To Deep Deterministic Policy Gradient, Aleksandr Svishchev

College of Graduate Studies: Theses & Dissertations

Reinforcement learning (RL) is a subfield of machine learning concerned with agents learning to behave optimally by interacting with an environment. One of the most important topics in RL is how the agent should explore, that is, how to choose actions in order to rate their impact on long-term reward. For example, a simple baseline strategy might be uniformly random action selection. This thesis investigates the heuristic idea that agents will learn faster if they explore by factoring the environment’s state into their decision and intentionally choose actions which are as different as possible from what they have previously observed. …

Classification In Supervised Statistical Learning With The New Weighted Newton-Raphson Method, Toma Debnath

College of Graduate Studies: Theses & Dissertations

In this thesis, the Weighted Newton-Raphson Method (WNRM), an innovative optimization technique, is introduced in statistical supervised learning for categorization and applied to a diabetes predictive model, to find maximum likelihood estimates. The iterative optimization method solves nonlinear systems of equations with singular Jacobian matrices and is a modification of the ordinary Newton-Raphson algorithm. The quadratic convergence of the WNRM, and high efficiency for optimizing nonlinear likelihood functions, whenever singularity in the Jacobians occur allow for an easy inclusion to classical categorization and generalized linear models such as the Logistic Regression model in supervised learning. The WNRM is thoroughly investigated …

Performing Holt-Winters Time Series Forecasting Using Neural Network Based Models, Kazeem Olanrewaju Bankole

College of Graduate Studies: Theses & Dissertations

We show how to create Artificial Neural Network based models for performing the well- known Holt-Winters time series analysis. Our work fares well compared to the well-known Holt-Winter time series prediction method while avoiding the burden of searching for the parameters of the model. We present the theoretical justification of the connection between the two models and experimental results showing the similarities of these models

Accounting For Variability Due To Resampling Using Bootstrapping, Dipendra Phuyal

College of Graduate Studies: Theses & Dissertations

Bradley Efron (1979) introduced bootrapping. Typically a researcher is interested in studying a process which generates individuals. The collection of individuals the process has(actual) or could have (conceptual) generated is the population. The collection of conceptual members of the population is an uncountable collection. Hence, the population is anuncountable collection of individuals. The collection of individuals the process has generated (actual individuals) is representative of what the process can generate and will bereferred to as the representative sample. The size of this sample is a nonnegative integervalued random variable N which may be a constant random variable such as in …

A Comparative Analysis Of A Family Of Advanced Iterative Optimization Methods In Nonlinear Regression, Tanmoy Kumar Debnath

College of Graduate Studies: Theses & Dissertations

Classical statistical supervised learning optimization techniques like the Gauss-Newton Iterative Method (GNIM), Weighted Gauss-Newton Iterative Method (WGNIM), Reweighted Gauss-Newton Iterative Method (RGNIM), and Levenberg-Marquart (LM) algorithm extend the nonlinear least squares method. The WGNIM improves model fitting by controlling heteroscedasticity in the linear and nonlinear models. A comparative analysis of the GNIM, WGNIM, RGNIM, and LM methods for fitting nonlinear models is presented. A step-wise diagnosis for structural multicollinearity in the reweighted linearized model is investigated via the Variance Inflation Factor (VIF) to determine variance inflation in the sequence of estimators for the model parameters. Under restricted multicollinearity levels in …

Cryptographic Algorithms, Cryptocurrencies, And A Predictive Model Of Bitcoin Value By Pls Regression, Paul Kenneth O'Connor

Masters Theses

"With the invention of Bitcoin in 2009, as a seemingly timed response to the ongoing financial crisis, the popularity of the cryptocurrency has since continued to grow. Just this year, the Security Exchange Commission approved Bitcoin for exchange traded funds, allowing major investment firms to begin product trading. With this approval, and during this very moment of writing, Bitcoin has entered a bull market and reached a record value of over 72,000 USD. In addition, the Bitcoin halving event in April of 2024 is expected to increase demand even further. It has been anticipated that Bitcoin and other cryptocurrencies will …

L∞ Bounds For Transient Growth In Repetitive And Iterative Learning Control Systems, Douglas A. Bristow, John R. Singler

Mechanical and Aerospace Engineering Faculty Research & Creative Works

This paper revisits the problem of large transient growth in Iterative Learning Control (ILC) and Repetitive Process Control (RPC) systems. In ILC and RPC problems a process is repeated iteratively, with new control calculations occurring in between each iteration. Large transient growth refers to the propensity of some control algorithms to grow error exponentially before eventually converging. While robust monotonic convergence algorithms (in which monotonic convergence is guaranteed usually in exchange for a small loss in performance) have largely eliminated the concern for large transient growth in ILC, similar results cannot always be obtained in RPC. The emergence of additive …

Utility In Time Description In Priority Best-Worst Discrete Choice Models: An Empirical Evaluation Using Flynn's Data, Sasanka Adikari, Norou Diawara

Mathematics & Statistics Faculty Publications

Discrete choice models (DCMs) are applied in many fields and in the statistical modelling of consumer behavior. This paper focuses on a form of choice experiment, best-worst scaling in discrete choice experiments (DCEs), and the transition probability of a choice of a consumer over time. The analysis was conducted by using simulated data (choice pairs) based on data from Flynn's (2007) 'Quality of Life Experiment'. Most of the traditional approaches assume the choice alternatives are mutually exclusive over time, which is a questionable assumption. We introduced a new copula-based model (CO-CUB) for the transition probability, which can handle the dependent …

A Copula Discretization Of Time Series-Type Model For Examining Climate Data, Dimuthu Fernando, Olivia Atutey, Norou Diawara

Mathematics & Statistics Faculty Publications

The study presents a comparative analysis of climate data under two scenarios: a Gaussian copula marginal regression model for count time series data and a copula-based bivariate count time series model. These models, built after comprehensive simulations, offer adaptable autocorrelation structures considering the daily average temperature and humidity data observed at a regional airport in Mobile, AL.

The Geometry Of Dynamic Time-Dependent Best-Worst Choice Pairs, Sasanka Adikari, Norou Diawara, Haim Bar

Mathematics & Statistics Faculty Publications

There has been increasing interest in best–worst discrete choice experiments (BWDCEs) in health economics, transportation research, and other fields over the last few years. BWDCEs have distinct advantages compared to other measurement approaches in discrete choice experiments (DCEs). A systematic study of best–worst (BW) choice pairs can be traced back to the 1990s. Recently, new ideas have been introduced to the subject. Calculating utility helps measure the attractiveness of BW choices. The goal of this paper is twofold. First, we extend the idea of the BW choice pair to include dynamic, time-dependent transition probability and capture utility at each time …

Innovative Solutions In Medical Diagnostics, Seafood Industry, Environmental, And Forensic Applications Through Integrating Raman Spectroscopy And Machine Learning, Luis Perez Almodovar

Electronic Theses & Dissertations (2024 - present)

There is an urgent need for rapid and on-site analytical methods to address various environmental and forensic challenges. This dissertation explores the application of Raman spectroscopy combined with machine learning for three critical areas: evaluating perfluoroalkyl or polyfluoroalkyl substances (PFAS) exposure in wildlife, authenticating tuna species in the seafood industry, identifying body fluids at crime scenes, and medical diagnostics.

First, Raman spectroscopy and support vector machine discriminant analysis (SVMDA) were employed to detect PFAS levels in the blood plasma of smallmouth bass from two different tributaries, rivers, or streams that flow into larger bodies of water. The method achieved 100% …

Advanced Techniques In Time Series Forecasting: From Deterministic Models To Deep Learning, Xue Bai

Graduate Theses, Dissertations, and Problem Reports (ETD)

This dissertation discusses three instances of temporal prediction, applied to population dynamics and deep learning.

In population modeling, dynamic processes are frequently represented by systems of differential equations, allowing for the analysis of various phenomena. The first application explores modeling cloned hematopoiesis in chronic myeloid leukemia (CML) via a nonlinear system of differential equations. By tracking the evolution of different cell compartments, including cycling and quiescent stem cells, progenitor cells, differentiated cells, and terminally differentiated cells, the model captures the transition from normal hematopoiesis to the chronic and accelerated-acute phases of CML. Three distinct non-zero steady states are identified, representing …

Title: I: L1-Norm Matrix Completion For Recommender Systems Ii: Conjecturing-Based Classification, Fatemeh Valizadeh Gamchi

Theses and Dissertations

Recommendation systems are essential for providing personalized user experiences, but their performance can be affected by outliers especially in traditional collaborative filtering methods that use the L2-norm. To address this challenge, we developed two new algorithms, SharpEl1rs and SharpEl1rs-Impute, based on the L1-norm to improve resistance against extreme values and effectively handle missing data. Our experimental setting was designed to compare these proposed methods with existing techniques. Then our algorithms are applied to real datasets to assess their performance, with findings indicating that our proposed models offer improved accuracy in some cases and solid performance in others for industrial-scale recommendation …

Redesign, Evaluation, And Validation Of A Commercially Viable High-Resolution Melt Based Mixture Screening Tool, Chastyn Smith

Theses and Dissertations

Analysis of evidentiary samples containing DNA from multiple contributors (“mixtures”) is a time intensive process for a forensic analyst and one where the contributor nature of a sample is not revealed until the end of the traditional forensic workflow. Often, at this stage, retesting or additional testing of mixture samples may not be possible, particularly if the DNA collection device did not preserve the DNA well enough; consequently leaving only trace amounts of a contributor’s DNA present. Thus, a new collection device that would allow for the increased preservation/integrity of evidentiary samples as well as a method that would allow …

Sparse Representation Learning For Temporal Networks, Maxwell Mcneil

Electronic Theses & Dissertations (2024 - present)

Temporal networks arise in many domains including activity of social network users, sensor network readings over time, and time course gene expression within the interaction network of a model organism. Data of this type contains a wealth of prior information such as the connectivity among nodes (e.g., a friendship graph), and prior knowledge of expected temporal patterns (e.g., periodicity). Modeling these temporal and network patterns jointly is essential for state-of-the-art performance in temporal network data analysis and mining. Sparse dictionary encoding is one modeling approach for such underlying patterns. However, most classical approaches consider only one dimension of the data …

Examining Stigma In Rural Mental Health Care Settings: A Mixed Methods Approach, Lainie Krumenacker

Murray State Theses and Dissertations

More than half of Americans will be diagnosed with a mental illness in their lifetime (CDC, 2021), yet stigma towards mental health affects both patients and providers. Although programs exist to address stigma, improve cultural competency among providers, and educate families on the importance of support, facilities are often limited on programs they provide due to allocation of resources and funds. Without a shift in treatment and programing, stigma will continue to impact patient care and outcome.

This study explored stigma among mental health providers in rural communities, while exploring potential differences in treatment among patients due to race. Mental …

Three Essays On Energy Related To State Policies And Low Carbon Transitions, Pinky Thomas

Graduate Theses, Dissertations, and Problem Reports (ETD)

This dissertation consists of three essays on energy-related state policies and energy transition. Each paragraph below refers to the three abstracts for the three chapters in this dissertation, respectively.

The first essay is entitled: “Impacts of State Tax and Resource Ownership Policies on Extraction: Evidence from U.S. Natural Gas Production”. The innovation of combined use of horizontal drilling and hydraulic fracturing technologies during the 2000s has allowed natural gas producers in the United States to extract natural gas and liquids from deep shale formations in a cost-efficient manner. This essay evaluates whether unconventional gas production responds to tax changes, and …

Investigation Of Space Charge Effects On Co2 Electrocatalytic Reduction On Gd-Doped Ceria Via Scanning Kelvin Probe And Model-Based Bayesian Analysis, Alejandro Mejia

Graduate Theses, Dissertations, and Problem Reports (ETD)

In studying novel energy conversion and storage systems, such as high-temperature electrolysis, numerous underlying fundamental physical processes remain unclear or inadequately understood. Among these, the modeling and comprehension of surface reaction mechanisms, coupled with the intricate effects of space‑charge interfaces, remains an unclear and challenging area of research.

The work of this dissertation involves the development of a 2D finite element analysis model, leveraging the robust MOOSE framework from INL. This model, featuring inhomogeneous defect thermodynamics for near-surface chemistry, formulated through Poisson‑Cahn variational theory, has been exploited for studying the electrocatalytic reduction of CO₂ on gadolinia doped ceria. The …

Data-Driven Modeling Of Oxygen Kinetics In La0.6sr0.4co0.2fe0.8o3−Δ (Lscf) For High-Temperature Reduction Of Co2 In An Electrolysis Cell, Ferron Campbell

Graduate Theses, Dissertations, and Problem Reports (ETD)

Electrolysis systems are critical to several societal applications, particularly energy storage and conversion. Developing these systems requires a detailed knowledge of the chemistry and thermodynamics of the materials used in the electrolysis cell. This work focuses on using embedded scientific machine learning as an efficient way to build an interpretable model for the reaction and transport kinetics in the LSCF electrode, whose performance directly influences the electrolysis system’s performance. The models developed in this study are trained using the publicly available machine learning package, FoKL-GP. This package incorporates a robust Gibbs sampler that employs a forward variable selection process to …

Extremal Graphs For Widom–Rowlinson Colorings In K-Chromatic Graphs, John Engbers, Aysel Erey

Mathematical and Statistical Science Faculty Research and Publications

The Widom–Rowlinson graph, H_WR , is the fully looped path on three vertices. Let hom(G,H_WR) be the number of graph homomorphisms from G to H_WR or, equivalently, the number of H_WR-colorings of G. We investigate extremal graphs for hom(G,H_WR) for G in the family of k-chromatic graphs subject to various connectivity requirements. In particular, we determine the graphs G maximizing hom(G,H_WR) in the families of n-vertex k-chromatic graphs, n-vertex connected k-chromatic graphs, n-vertex k-chromatic graphs with c components, n …

A Journey From Univariate To Multivariate Functional Time Series: A Comprehensive Review, Hossein Haghbin, Mehdi Maadooliat

Mathematical and Statistical Science Faculty Research and Publications

Functional time series (FTS) analysis has emerged as a potent framework for modeling and forecasting time-dependent data with functional attributes. In this comprehensive review, we navigate through the intricate landscape of FTS methodologies, meticulously surveying the core principles of univariate FTS and delving into the nuances of multivariate FTS. The journey commences with an exploration of the foundational aspects of univariate FTS analysis. We delve into representation, estimation, and modeling, spotlighting the effectiveness of various parametric and nonparametric models at our disposal. The stage then transitions to multivariate FTS analysis, where we confront the intricacies posed by high-dimensional data. We …

Gwid: An R Package And Shiny Application For Genome-Wide Analysis Of Ibd Data, Soroush Mahmoudiandehkordi, Mehdi Maadooliat, Steven J. Schrodi

Mathematical and Statistical Science Faculty Research and Publications

Summary

Genome-wide identity by descent (gwid) is an R package developed for the analysis of identity-by-descent (IBD) data pertaining to dichotomous traits. This package offers a set of tools to assess differential IBD levels for the two states of a binary trait, yielding informative and meaningful results. Furthermore, it provides convenient functions to visualize the outcomes of these analyses, enhancing the interpretability and accessibility of the results. To assess the performance of the package, we conducted an evaluation using real genotype data derived from the SNPs to investigate rheumatoid arthritis susceptibility from the Marshfield Clinic Personalized Medicine Research Project.

Availability …

Maximizing The Number Of H-Colorings Of Graphs With A Fixed Minimum Degree, John Engbers

Mathematical and Statistical Science Faculty Research and Publications

For graphs G and H, an H-coloring of G is an adjacency-preserving map from the vertex set of G to the vertex set of H.

Development Of Probabilistic Dynamic Model Building And Bayesian Machine Learning Approaches, Samuel Oladayo Adeyemo

Graduate Theses, Dissertations, and Problem Reports (ETD)

Abstract

Development of Probabilistic Dynamic Model Building and Bayesian Machine Learning Approaches

Samuel Adeyemo

The recent years have seen a tremendous increase in the use of artificial intelligence (AI) and machine learning (ML) for the development of data-driven mathematical models needed for performing real-time optimization, model-based control, performance optimization, dynamic data reconciliation, and process performance monitoring. However, the development of data-driven models is faced with some challenges including lack of model interpretability, sensitivity of algorithm to noise in training data, limited extrapolation capabilities and violation of conservation laws. Drawing motivation from these existing gaps, this work aims to develop robust …