Master of Science, University of Toledo, 2020, Engineering (Computer Science)

In many application settings involving networks, such as friendships or messages among users of an on-line social network and transactions between traders in financial markets, understanding network dynamics has been a long-standing problem with implications in numerous disciplines including computer science, physics, mathematics, biology, social sciences, and economics. Due to high computational complexity of dynamic network analysis, most models assume networks are static which sacrifices expressiveness for scalability. Here we set forth two new concepts which enhance link prediction and recommendation as well as modeling communities and interactions in continuous-time dynamic networks. In particular, we first introduce the notion of personalized degree and find that neighbors with higher personalized degree are more likely to lead to new link formations when they serve as common neighbors with other nodes, both in undirected and directed settings. Next, we propose the Community Hawkes Independent Pairs (CHIP) generative model for continuous-time networks of timestamped relational events. We show that spectral clustering provides consistent community detection, for a growing number of nodes, on networks generated by the CHIP model and develop consistent and computationally efficient estimators for the model parameters. Personalized degree provides a lens into the latent information in the topology of on-line social networks and how this information can be utilized to better understand the evolution of a network over time and to predict future interactions. We find that incorporating personalized degree into common neighbor based link prediction algorithms can improve mean link prediction accuracy by up to 35%, while incorporating directions of edges further improves accuracy by up to 11%. Moreover, the CHIP model is able to capture network dynamics and its underlying community structure with only a few parameters and scale to much larger networks compared t (open full item for complete abstract)

Committee: Kevin Xu (Committee Chair); Ahmad Y Javaid (Committee Member); Qin Shao (Committee Member) Subjects: Computer Science

Doctor of Philosophy (Ph.D.), Bowling Green State University, 2024, Statistics

Deep learning, neural network, has been penetrating into almost every corner of data analyses. With advantages on computing power and speed, adding more layers in a neural network analysis becomes a common practice to improve the prediction accuracy. However, over depleting information in the training dataset may consequently carry data noises into the learning process of neural network and result in over-fitting errors. Neural Network has been used to predict the future time series data. It had been claimed by several authors that the Neural Network (Recurrent Neural Network) can predict the time series data, although time series models have also been used to predict the future time series data. This dissertation is thus motivated to investigate the prediction performances of neural networks versus the conventional inference method of time series analysis. After introducing basic concepts and theoretical background of neural network and time series prediction in Chapter 1, Chapter 2 analyzes fundamental structure of time series, along with estimation, hypothesis testing, and prediction methods. Chapter 3 discusses details of computing algorithms and procedures in neural network with theoretical adjustment for time series prediction. In conjunction with terminologies and methodologies in the previous chapters, Chapter 4 directly compares the prediction results of neural networks and conventional time series for the squared error function. In terms of methodology assessment, the evaluation criterion plays a critical role. The performance of the existing neural network models for time series predictions has been observed. It has been experimentally observed that the time series predictions by time series models are better compared to the neural network models both computationally and theoretically. The conditions for the better performances of the Time Series Models over the Neural Network Models have been discovered. Theorems have also been pro (open full item for complete abstract)

Committee: John Chen Ph.D. (Committee Chair); Hanfeng Chen Ph.D. (Committee Member); Umar Islambekov Ph.D. (Committee Member); Brigid Burke Ph.D. (Other) Subjects: Applied Mathematics; Artificial Intelligence; Behavioral Sciences; Computer Science; Education Finance; Finance; Information Systems; Operations Research; Statistics

Doctor of Philosophy, University of Toledo, 2021, Engineering

The importance of modeling the spreading processes through a population has led to the development of several mathematical models. A number of empirical studies have collected and analyzed data on contacts between individuals which also shows patterns of contacts in an ever-evolving network. Contagious processes on networks, such as the spread of disease through physical proximity or information diffusion over social media, are continuous-time processes that depend upon the pattern of interactions between the individuals in the network. Continuous-time stochastic epidemic models are a natural fit for modeling the dynamics of such processes. However, prior works on such continuous-time models do not consider the dynamics of the underlying interaction network which involves the addition and removal of edges over time. In this work, firstly, we investigate the effects of different contact network models with varying levels of complexity on the outcomes of simulated epidemics. Secondly, we incorporate continuous-time network dynamics (addition and removal of edges) into continuous-time epidemic simulations and propose two rejection-sampling based approaches coupled with the well-known Gillespie algorithm and Thinning algorithm that enables exact simulation of the continuous-time epidemic process. Thirdly, we propose a continuous-time contact network model which takes into account the duration of contacts for inference procedure.

Committee: Kevin Xu (Committee Chair); Devinder Kaur (Committee Member); Rong Liu (Committee Member); Ahmad Javaid (Committee Member); Defne Apul (Committee Member); Ezzatollah Salari (Committee Member) Subjects: Computer Science; Epidemiology; Mathematics; Sociology; Statistics

Doctor of Philosophy (Ph.D.), University of Dayton, 2018, Mechanical Engineering

This research presents engineering models that simulate steady-state and transient operations of air-cooled condensing units and an automatic commercial ice making machines ACIM, respectively. The models use easily-obtainable inputs and strategies that promote quick computations. Packaged, air-cooled condensing units include a compressor, condensing coil, tubing, and fans, fastened to a base or installed within an enclosure. A steady-state standard condensing unit system simulation model is assembled from conventional, physics-based component equations. Specifically, a four-section, lumped-parameter approach is used to represent the condenser, while well-established equations model compressor mass flow and power. To increase capacity and efficiency, enhanced condensing units include an economizer loop, configured in either upstream or downstream extraction schemes. The economizer loop uses an injection valve, brazed-plate heat exchanger (BPHE) and scroll compressor adapted for vapor injection. An artificial neural network is used to simulate the performance of the BPHE, as physics-based equations provided insufficient accuracy. The capacity and power results from the condensing unit model are generally within 5% when compared to the experimental data. A transient ice machine model calculates time-varying changes in the system properties and aggregates performance results as a function of machine capacity and environmental conditions. Rapid "what if" analyses can be readily completed, enabling engineers to quickly evaluate the impact of a variety of system design options, including the size of the air-cooled heat exchanger, finned surfaces, air flow rate, ambient air and inlet water temperatures, compressor capacity and/or efficiency for freeze and harvest modes, refrigerants, suction/liquid line heat exchanger and thermal expansion valve properties. Simulation results from the ACIM model were compared with the experimental data of a fully instrumented, standar (open full item for complete abstract)

Committee: David Myszka (Advisor); Kevin Hallinan (Committee Member); Andrew Chiasson (Committee Member); Rajan Rajendran (Committee Member) Subjects: Computer Engineering; Condensation; Conservation; Design; Endocrinology; Energy; Engineering; Environmental Economics; Environmental Education; Environmental Engineering; Environmental Science; Mechanical Engineering

Master of Science, The Ohio State University, 2018, Mechanical Engineering

Thermal management of aftertreatment devices is nowadays becoming a critical requirement to comply with the stringent emission standards. A recent challenge in aftertreatment control and diagnostics is the need to monitor the temperature distribution along the entire length of the catalyst, rather than relying on a single-point measurement. To this extent, this work focuses on the development and validation of a control-oriented, physics-based Three Way Catalytic Converter (TWCC) model for the purpose of real-time thermal monitoring. Starting from the governing equations in Partial Dierential Equation (PDE) form, a model order reduction technique that combines Proper Orthogonal Decomposition and Collocation is developed. The sensitivity of the selection of the empirical basis functions is studied. To include the exothermic effect of chemical reactions, an Articial Neural Network is trained. To study the accuracy of the reduced order model, two high-fidelity models using Finite Dierence Method and commercial simulation software GT-Power are developed. The reduced order model executes much faster than using standard numerical methods, meanwhile providing comparable accuracy. The model is validated against GT-POWER in the ability to predict the temperature distribution, the peak temperature and its axial location in US-06 Drive Cycle.

Committee: Marcello Canova (Advisor); Shawn Midlam-Mohler (Committee Member) Subjects: Mechanical Engineering

Master of Science, The Ohio State University, 2018, Computer Science and Engineering

In recent decades, there has been an increasing interest in the capability of a robot follower for both human and robot leaders. In the case of a human leader, there is sometimes a requirement of additional sensors, GPS tracking, or WIFI capability. While these solutions may produce the desired results, they are often unrealistic for industry, particularly when using off-the-shelf hardware, or when WIFI or GPS is not available. This work describes an industry-friendly robot follower, which utilizes only a mobile device, camera, and Deep Neural Network (DNN) to keep pace with a designated human leader. We do this by collecting inertial data from the leader's mobile device, which is then transferred to the robot's server via Bluetooth. This data is combined with a corresponding video from the robot's camera, which is then analyzed using a DNN to classify the action the robot should take. These components in tandem result in a robust following system, which can be minutely tweaked for a variety of scenarios and requirements.

Committee: Dong Xuan (Advisor); Feng Qin (Committee Member) Subjects: Computer Science

Master of Science (MS), Ohio University, 2017, Industrial and Systems Engineering (Engineering and Technology)

Weather conditions affect road traffic in many ways such as traffic demand, traffic safety, and traffic flow. The influence of the weather events might cause extra cost and time if they are not considered during travel plan. The efficiency of the travel activities under different road conditions can be improved in terms of safety, mobility, and travel time saving. The aim of this study is to improve travel planning performance measures under different weather conditions. Vehicle speed data from weigh in motion traffic recorder (WIM) along with weather data have been used to obtain the vehicle speed reduction during adverse weather events. Also, vehicle speed during the events were compared with the normal vehicle speed under regular weather condition for different locations in the State of Ohio, and time durations. Results showed that heavy snow associated with low visibility have the most impacts on driver speed behavior. Then, the vehicle speed performance during the events was used to calculate the expected time throughout each event. Using probability approach on two different networks show significant results in terms of departure times and route choices. Fuzzy mathematical model was conducted to determine if there are significant differences in the satisfaction levels compared to probability approach when the snow storm hits all the links in the network. Thus, the results showed that there is not remarkable difference between the fuzzy and probability. This methodology can be used to demonstrate a predictive tool to assess departure time and route choice under snow storm events. These approaches may help decision makers to obtain the optimal path before and during adverse weather events. This concept is presented in the context of the transportation network planning performance.

Committee: Gursel Suer (Advisor); Diana Schwerha (Committee Member); Tao Yuan (Committee Member); Mohammed Bhutta (Committee Member) Subjects: Civil Engineering; Industrial Engineering; Statistics; Transportation; Transportation Planning

PhD, University of Cincinnati, 2010, Engineering : Mechanical Engineering

Simulation-based study of human auditory response characteristics and development of a prototype for advanced noise guideline are two major focuses of this dissertation research. This research was conducted as a part of the long-term effort to develop an improved noise guideline for better protection of the workers exposed to various noise environments. The human auditory responses were studied with simulation models. A human full-ear model derived from an existing model, Auditory Hazard Assessment Algorithm for Human (AHAAH), was utilized as a baseline for the study. Frequency- and time-domain responses of well-known human middle ear network models were cross-compared to estimate expected accuracy of the models and understand their proper use. Responses of the stapes to impulsive noises were investigated by using the middle ear models to understand the effects of the temporal characteristics of impulsive noises on the responses. Available measured transfer functions between the free-field pressure and the stapes response for human and chinchilla were also used to study the auditory response characteristics. The measured transfer functions were refined and reconditioned to make them have equivalent formats. Using the reconstructed transfer functions, time-domain stapes responses of human and chinchilla to impulsive and complex type noises were calculated and compared. Applicability of the noise metrics defined in terms of the stapes response to assess the risk of the noise induced hearing loss was studied. A prototype of an improved noise guideline was developed from existing chinchilla noise exposure data. Applying a new signal processing technique to the time histories of the exposed noises and studying the relationship between the noise metric and the permanent threshold shift (PTS), the dose-response relationship (DRR) was established in a compatible way with the definition used in current human noise guidelines. From the DDR, noise induced hearing loss (NIHL (open full item for complete abstract)

Committee: J. Kim PhD (Committee Chair); William Murphy PhD (Committee Member); Mark Schulz PhD (Committee Member); Teik Lim PhD (Committee Member) Subjects: Mechanical Engineering

Doctor of Philosophy, The Ohio State University, 2008, Computer and Information Science

This work seeks to develop a probabilistic framework for modeling, querying and analyzing large-scale structured and semi-structured data. The framework has three components: (1) Mining non-redundant local patterns from data; (2) Gluing these local patterns together by employing probabilistic models (e.g., Markov random field (MRF), Bayesian network); and (3) Reasoning over the data for solving various data analysis tasks. Our contributions are as follows: (a) We present an approach of employing probabilistic models to identify non-redundant itemset patterns from a large collection of frequent itemsets on transactional data. Our approach can effectively eliminate redundancies from a large collection of itemset patterns. (b) We propose a technique of employing local probabilistic models to glue non-redundant itemset patterns together in tackling the link prediction task in co-authorship network analysis. Our technique effectively combines topology analysis on network structure data and frequency analysis on network event log data. The main idea is to consider the co-occurrence probability of two end nodes associated with a candidate link. We propose a method of building MRFs over local data regions to compute this co-occurrence probability. Experimental results demonstrate that the co-occurrence probability inferred from the local probabilistic models is very useful for link prediction. (c) We explore employing global models, models over large data regions, to glue non-redundant itemset patterns together. We investigate learning approximate global MRFs on large transactional data and propose a divide-and-conquer style modeling approach. Empirical study shows that the models are effective in modeling the data and approximately answering queries on the data. (d) We propose a technique of identifying non-redundant tree patterns from a large collection of structural tree patterns on semi-structured XML data. Our approach can effectively eliminate redundancies from a larg (open full item for complete abstract)

Committee: Srinivasan Parthasarathy (Advisor) Subjects: Computer Science

Doctor of Philosophy, The Ohio State University, 2004, Geodetic Science and Surveying

Reliability analysis explains the contribution of each observation in an estimation model to the overall redundancy of the model, taking into account the geometry of the network as well as the precision of the observations themselves. It is principally used to design networks resistant to outliers in the observations by making the outliers more detectible using standard statistical tests.It has been studied extensively, and principally, in Gauss-Markov models. We show how the same analysis may be extended to various constrained Gauss-Markov models and present preliminary work for its use in unconstrained Gauss-Helmert models. In particular, we analyze the prominent reliability matrix of the constrained model to separate the contribution of the constraints to the redundancy of the observations from the observations themselves. In addition, we make extensive use of matrix differential calculus to find the Jacobian of the reliability matrix with respect to the parameters that define the network through both the original design and constraint matrices. The resulting Jacobian matrix reveals the sensitivity of reliability matrix elements highlighting weak areas in the network where changes in observations may result in unreliable observations. We apply the analytical framework to photogrammetric networks in which exterior orientation parameters are directly observed by GPS/INS systems. Tie-point observations provide some redundancy and even a few collinear tie-point and tie-point distance constraints improve the reliability of these direct observations by as much as 33%. Using the same theory we compare networks in which tie-points are observed on multiple images (n-fold points) and tie-points are observed in photo pairs only (two-fold points). Apparently, the use of two-fold tie-points does not significantly degrade the reliability of the direct exterior observation observations. Coplanarity constraints added to the common two-fold points do not add significantly to the (open full item for complete abstract)

Committee: Burkhard Schaffrin (Advisor) Subjects: Engineering, Civil; Geodesy

Master of Science in Electrical Engineering (MSEE), Wright State University, 2024, Electrical Engineering

Performing automatic target recognition (ATR) on full-size aircraft targets using inverse synthetic aperture radar (ISAR) data is challenging and expensive. The use of scale models and radar systems of such large targets saves time and reduces facility requirements. This study examines the feasibility of performing ATR on 1:144 scale model airplanes at Ka-band. The scale model and Ka-band radar simulate the collection of full-scale targets at VHF-band. The phase history measurement collections were completed in the Sensors and Signals Exploitation Laboratory (SSEL) at Wright State University. To ensure sufficient data for training and testing, the phase history data was augmented through mathematical translation and rotation of the scene. These augmented images were processed using the polar format algorithm and subsequently classified using support vector machines and convolutional neural networks. The resulting ATR models achieved a classification accuracy of over 82 percent for all aircraft types, except for the very similar B747-8 and B747-8F, which exhibited misclassification rates consistent with expectations for such similar targets.

Committee: Michael A. Saville Ph.D. (Advisor); Cheryl B. Schrader Ph.D. (Committee Member); Michael L. Raymer Ph.D. (Committee Member); Josh Ash Ph.D. (Committee Member) Subjects: Electrical Engineering

Doctor of Philosophy, Case Western Reserve University, 0, EECS - Computer and Information Sciences

In recent years, there has been a significant increase in our capacity to collect data from online platforms or physical sensors. Networks often serve as a universal representation for relational data such as online social networks, biology, transportation systems, and chemical structures. My work in this dissertation focuses on developing probabilistic network models to offer a structured framework for representing, inferring and learning uncertainties in complex network data. The first segment focuses on health informatics, specifically developing biological network models for kidney transplants to better predict survival outcomes. I construct a latent space model for HLA networks, an unobserved bipartite network, to enhance survival analysis. The second part of my research delves into the area of social network analysis. In real world, most of social networks are dynamic. Due to the significant computational demands of analyzing dynamic networks, traditional approaches often simplify networks to static representation, losing some dynamic detail for computational ease. My aim is to better capture these dynamics. To this end, I introduce continuous-time network models that leverage a latent space approach and temporal point processes, significantly enhancing the analysis and prediction capabilities in dynamic network studies. The third part builds upon the first work by incorporating 'marks' information in continuous-time networks, representing connection durations between node pairs. This section involves creating probabilistic network models for interval networks, extending the predictive capabilities of the earlier models. The main objective for my work not only enhances the predictive capabilities of machine learning models within networks but also provides valuable insights into the underlying structural and dynamic properties of interconnected entities.

Committee: Kevin Xu (Advisor); Mehmet Koyutürk (Committee Member); Subhadeep Paul (Committee Member); Abhinendra Singh (Committee Member); Sanmukh Kuppannagari (Committee Member) Subjects: Computer Science

Master of Computing and Information Systems, Youngstown State University, 2024, Department of Computer Science and Information Systems

The field of coding education is rapidly evolving, with emerging technologies playing a pivotal role in transforming traditional learning methodologies. This thesis introduces Dr. Lego, an innovative framework designed to revolutionize the assessment and understanding of block-based coding through the integration of sophisticated deep learning models. Dr. Lego combines cutting-edge technologies such as MobileNetV3 (Howard, 2019), for visual recognition and BERT (Devlin et al., 2018), and XLNet (Yang et al., 2019) for natural language processing to offer a comprehensive approach to evaluating coding proficiency. The research methodology involves the meticulous curation of a diverse dataset comprising projects from the LEGO SPIKE app (LEGO Education, 2022), ensuring that the models are subjected to a broad range of coding scenarios. Leveraging the dynamic educational environment provided by the LEGO SPIKE app (LEGO Education, 2022), Dr. Lego empowers users to design and implement various coding projects, fostering hands-on learning experiences. This thesis delves into methodologies aimed at enhancing coding education by exploring model integration, data generation, and fine-tuning of pre-trained models. Dr. Lego not only evaluates coding proficiency but also provides cohesive and insightful feedback, enhancing the learning experience for users. The adaptability of the framework highlights its potential to shape the future of coding education, paving the way for a new era of interactive and engaging learning experiences.

Committee: Abdu Arslanyilmaz PhD (Advisor); Feng Yu PhD (Committee Member); Carrie Jackson EdD, BCBA (Committee Member) Subjects: Computer Science; Engineering; Information Systems; Robotics; Teaching

Master of Arts, Miami University, 2024, Economics

This study attempts to estimate direct network effects within the online video game platform Steam. To overcome Manski's ``reflection problem" in identifying this effect, I use three holidays - the Chinese National Day in 2023, the American Thanksgiving in 2023, and the American Martin Luther King Holiday in 2024 - as exogenous regional shocks and employ the differences-in-differences method on time and players' location. Using data from Steam API, I estimate a model of daily playtime and the choice of whether to play at all, allowing utility to vary with players' own location, day of the week, and their friends' locations. This model permits the localized nature of network effects observed in the data. However, because of the limited number of observed US-to-non-US and Chinese-to-non-Chinese friendships on the platform, the data is underpowered to detect an effect.

Committee: Charles Moul (Advisor); Josh Ederington (Committee Member); Peter Nencka (Committee Member) Subjects: Economics

Doctor of Philosophy (PhD), Wright State University, 2023, Engineering PhD

The design of Hypersonic Vehicles (HVs) requires meeting multiple unconventional and often conflicting design requirements in a hostile, high-energy environment. The most fundamental difference between ordinary aerospace design and hypersonic flight is that the extreme conditions of hypersonic flight require parts to perform multiple functions and be tightly integrated, resulting in significant coupled effects. Critical couplings among the disciplines of aerodynamics, structures, propulsion, and thermodynamics must be investigated in the early stages of design exploration to reduce the risk of requiring major design changes and cost overruns later. In addition, due to a lack of validated test data within the coupled high-dimensional design domains, concept design exploration of HVs poses unprecedented challenges, especially in terms of computational costs and decision-making under uncertainty. A common design exploration technique is to sample the expensive physics-based models in a design of experiments and then use the sample data to train an inexpensive metamodel. Conventional metamodels include Polynomial Chaos Expansion, kriging, and neural networks. However, many simulation evaluations are needed for the design of experiments because of the large number of independent parameters for each design and the complex responses resulting from interactions across multiple disciplines. Because each simulation is expensive, the total costs are often computationally intractable. Computational cost reduction is often achieved using Multi-Fidelity (MF) modeling and Active Learning (AL). MF models supplement High-Fidelity (HF) simulations with less accurate but inexpensive Low-Fidelity (LF) simulations. AL generates training data in an iterative process: rebuilding the metamodel after each HF sample is added, and then using the metamodel to select the next HF sample. Location-specific uncertainty information is critical for making this determination. To address t (open full item for complete abstract)

Committee: Harok Bae Ph.D. (Advisor); Edwin Forster Ph.D. (Committee Member); Ramana Grandhi Ph.D. (Committee Member); Mitch Wolff Ph.D. (Committee Member); Nathan Klingbeil Ph.D. (Committee Member); Jose Camberos Ph.D. (Committee Member) Subjects: Aerospace Engineering; Artificial Intelligence; Engineering; Mechanical Engineering; Statistics

Master of Science, Miami University, 2023, Computer Science and Software Engineering

Software models are used to analyze and understand the properties of the system, providing stakeholders with an overview of how the system should work before actually implementing it. Such models are usually created informally, such as drawing sketches on a whiteboard or paper, especially during the early design phase, because these methods foster communication and collaboration among stakeholders. However, these informal sketches must be formalized to be useful in later applications, such as analysis, code generation, and documentation. This formalization process is often tedious, error-prone, and time-consuming. In an effort to avoid recreating formal models from scratch, this thesis presents SkeMo, a sketch-based software modeling tool. SkeMo is built on a CNN-based image classifier using 3000 input sketches of class diagram components and integrated into the functionality of an existing web-based model editor, the Instructional Modeling Language (IML), with a newly implemented touch interface. SkeMo was evaluated using a ten-fold cross-validation to assess the image classifier and through a user study involving 20 participants to collect metrics and feedback. The results demonstrate the promising potential of sketch-based modeling as an intuitive and efficient modeling practice, allowing users to quickly and easily create models to design complex software systems.

Committee: Eric Rapos (Advisor); Christopher Vendome (Committee Member); Xianglong Feng (Committee Member); Douglas Troy (Committee Member) Subjects: Computer Science; Engineering

MS, University of Cincinnati, 2022, Engineering and Applied Science: Electrical Engineering

The ever-changing and rapid advances in computer vision, machine learning has paved way to unparalleled likelihoods in the broad world of sports, most especially in soccer. In current times, many professional sports teams use artificial intelligence and machine learning to improve the training and develop better strategies for competition. Great emphasis has been placed on collecting data, implementing models and techniques that help simplify set tasks by the teams. In this thesis, we developed an effective way of applying image and video processing techniques, deep learning, and machine learning models to predict penalty kicks in a soccer game. Specifically, the deep neural networks are used to analyze soccer videos and to predict penalty kicks' trajectories. The image and video enhancement techniques are used to adjust the pixel values in the videos and improve the performance of the model which reduces the average and final displacement errors associated with predicting the trajectories of the player run up and the final kick. We conclude by extracting the visual and semantic features of the videos and testing with a multi-task prediction model in order to achieve a better result of trajectory prediction.

Committee: Wen-Ben Jone Ph.D. (Committee Member); Xuefu Zhou Ph.D. (Committee Member) Subjects: Engineering

Doctor of Philosophy (Ph.D.), University of Dayton, 2021, Mechanical Engineering

Indoor thermal comfort in residential buildings is usually achieved by tenants manually adjusting fixed temperature set-points; this is known as a ‘static' method. Prior research has explored automated control of thermal comfort based on the concept of a Predicted Mean Vote (PMV) index, which has been developed to provide a model of perceived human comfort. However, one of the dominant contributions to this index, the Mean Radiant Temperature (MRT), effectively the mean radiant temperature of the surrounding interior surfaces, has either been: 1) inaccurately assumed to be the same as indoor air temperature; and/or 2) costly to implement due to the need for numerous additional sensors. Research is posed to leverage prior work in automatically estimating the R-values of walls and ceilings using a combination of smart WiFi thermostat, building geometry, and historical energy consumption [51] to estimate the MRT with accuracy and thus provide a means to control for comfort, rather than temperature alone. In order to assess the energy saving potential of comfort control for any residence, a machine learning model of the indoor temperature based upon a NARX Neural Network is employed. This model leverages historical thermostat and weather data to develop a means to dynamically predict the interior temperature. With a developed model, it is possible to simulate different temperature set-points on indoor temperature, and thus identify the optimal set-point temperature at all times needed to maintain a reasonable comfort condition. Application of this ideal temperature set-point for minimum human comfort to historical weather data and indoor weather conditions can yield an estimate for minimum cooling energy. The initial results showed cooling energy savings in excess of 83% and 95%, respectively, for high- and low-efficiency residences. Based on this research, it is proposed that the approach to estimate MRT can be used to calculate a more accurate PMV value and a better r (open full item for complete abstract)

Committee: Timothy Reissman (Committee Chair); Rajen Rajendran (Committee Member); Andrew Chiasson (Committee Member); Kevin Hallinan (Committee Co-Chair) Subjects: Energy; Mechanical Engineering

Doctor of Philosophy, Case Western Reserve University, 2022, Physics

In the last two decades, advances in experimental techniques have opened up new vistas for understanding bio-molecules and their complex networks of interactions in the cell. In this thesis, we use theoretical modeling and machine learning to explore two surprising aspects that have been revealed by recent experiments: (i) the discovery that many different types of cellular signaling networks, in both prokaryotes and eukaryotes, can transmit at most 1 to 3 bits of information; (ii) the observation that single bio-molecules can exhibit multiple, stable conformational states with extremely heterogeneous functional properties. The first part of the thesis investigates how the energetic costs of signaling in biological networks constrain the amount of information that can be transferred through them. The focus is specifically on the kinase-phosphatase enzymatic network, one of the basic elements of cellular signaling pathways. We find a remarkably simple analytical relationship for the minimum rate of ATP consumption necessary to achieve a certain signal fidelity across a range of frequencies. This defines a fundamental performance limit for such enzymatic systems, and we find evidence that a component of the yeast osmotic shock pathway may be close to this optimality line. By quantifying the evolutionary pressures that operate on these networks, we argue that this is not a coincidence: natural selection is capable of pushing signaling systems toward optimality, particularly in unicellular organisms. Our theoretical framework is directly verifiable using existing experimental techniques, and predicts that many more examples of such optimality should exist in nature. In the second part of the thesis, we develop two machine learning methods to analyze data from single-molecule AFM pulling experiments: a supervised (deep learning) and an unsupervised (non-parametric Bayesian) algorithm. These experiments involve applying an increasing force on a bio-molecul (open full item for complete abstract)

Committee: Michael Hinczewski (Committee Chair); Peter Thomas (Committee Member); Harsh Mathur (Committee Member); Lydia Kisley (Committee Member) Subjects: Biophysics; Physics

Master of Science, Miami University, 2020, Computer Science and Software Engineering

Online discussion forums are increasingly used in large classrooms to enhance students' collaboration, promote student engagement and measure academic performances. Although many Computer-Supported Collaborative Learning (CSCL) tools are available, only a few researchers in the past have investigated the limitations of unstructured discussion forums. The linearly structured threads of discussion posts in these discussion forums make it difficult to represent and analyze the student interaction patterns. Modeling these discussion forums as collaboration networks enables the analysis of students' interaction and learning behavior. This thesis work demonstrates the need for and value of innovative network models that involve developing more structured, schema-based, and goal-oriented discussion forums constructed from active student collaborations. The schema-based structured network model enables identifying and measuring the influence that the student interactions have in dynamically evolving student collaboration networks as the course progresses. First, the unstructured discussion boards in Canvas are analyzed using graph mining and social network techniques. Then, the structured collaborative network model is used to study the impact that student collaborations have on knowledge acquisition, persistence and course outcomes using machine learning algorithms. These structured discussions enable the analysis of the changes in student collaboration patterns and belongingness for pedagogical benefits.

Committee: Vijayalakshmi Ramasamy (Advisor); James D. Kiper (Committee Member); Hakam W. Alomari (Committee Member) Subjects: Computer Science; Education