Theses and Dissertations
This collection includes both ASU Theses and Dissertations, submitted by graduate students, and the Barrett, Honors College theses submitted by undergraduate students.
Displaying 21 - 30 of 99
Description
Generative models are deep neural network-based models trained to learn the underlying distribution of a dataset. Once trained, these models can be used to sample novel data points from this distribution. Their impressive capabilities have been manifested in various generative tasks, encompassing areas like image-to-image translation, style transfer, image editing, and more. One notable application of generative models is data augmentation, aimed at expanding and diversifying the training dataset to augment the performance of deep learning models for a downstream task. Generative models can be used to create new samples similar to the original data but with different variations and properties that are difficult to capture with traditional data augmentation techniques. However, the quality, diversity, and controllability of the shape and structure of the generated samples from these models are often directly proportional to the size and diversity of the training dataset. A more extensive and diverse training dataset allows the generative model to capture overall structures present in the data and generate more diverse and realistic-looking samples.
In this dissertation, I present innovative methods designed to enhance the robustness and controllability of generative models, drawing upon physics-based, probabilistic, and geometric techniques. These methods help improve the generalization and controllability of the generative model without necessarily relying on large training datasets. I enhance the robustness of generative models by integrating classical geometric moments for shape awareness and minimizing trainable parameters. Additionally, I employ non-parametric priors for the generative model's latent space through basic probability and optimization methods to improve the fidelity of interpolated images. I adopt a hybrid approach to address domain-specific challenges with limited data and controllability, combining physics-based rendering with generative models for more realistic results.
These approaches are particularly relevant in industrial settings, where the training datasets are small and class imbalance is common. Through extensive experiments on various datasets, I demonstrate the effectiveness of the proposed methods over conventional approaches.
ContributorsSingh, Rajhans (Author) / Turaga, Pavan (Thesis advisor) / Jayasuriya, Suren (Committee member) / Berisha, Visar (Committee member) / Fazli, Pooyan (Committee member) / Arizona State University (Publisher)
Created2023
Description
This dissertation centers on the development of Bayesian methods for learning differ- ent types of variation in switching nonlinear gene regulatory networks (GRNs). A new nonlinear and dynamic multivariate GRN model is introduced to account for different sources of variability in GRNs. The new model is aimed at more precisely capturing the complexity of GRN interactions through the introduction of time-varying kinetic order parameters, while allowing for variability in multiple model parameters. This model is used as the drift function in the development of several stochastic GRN mod- els based on Langevin dynamics. Six models are introduced which capture intrinsic and extrinsic noise in GRNs, thereby providing a full characterization of a stochastic regulatory system. A Bayesian hierarchical approach is developed for learning the Langevin model which best describes the noise dynamics at each time step. The trajectory of the state, which are the gene expression values, as well as the indicator corresponding to the correct noise model are estimated via sequential Monte Carlo (SMC) with a high degree of accuracy. To address the problem of time-varying regulatory interactions, a Bayesian hierarchical model is introduced for learning variation in switching GRN architectures with unknown measurement noise covariance. The trajectory of the state and the indicator corresponding to the network configuration at each time point are estimated using SMC. This work is extended to a fully Bayesian hierarchical model to account for uncertainty in the process noise covariance associated with each network architecture. An SMC algorithm with local Gibbs sampling is developed to estimate the trajectory of the state and the indicator correspond- ing to the network configuration at each time point with a high degree of accuracy. The results demonstrate the efficacy of Bayesian methods for learning information in switching nonlinear GRNs.
ContributorsVélez-Cruz, Nayely (Author) / Papandreou-Suppappola, Antonia (Thesis advisor) / Moraffah, Bahman (Committee member) / Tepedelenlioğlu, Cihan (Committee member) / Berisha, Visar (Committee member) / Arizona State University (Publisher)
Created2023
Description
Trace evidence is an essential component of forensic investigations. Anthropogenicmaterials such as fibers and glass have been well studied for use in forensic trace evidence, but the potential use of retroreflective beads found in soils for forensic investigations is largely unexplored. Retroreflective glass beads are tiny spheres mixed into pavement markings to create reflective surfaces to reduce lane departure accidents. Retroreflective glass beads are a potentially new source of trace evidence for forensic investigations.
Analysis of the spatial distribution and chemical compositions of retroreflective glass beads recovered from 17 soil samples were analyzed and compared to see if there are striking variations that can distinguish samples by source. Soil samples taken near marked roads showed significantly higher concentrations of glass beads, averaging from 0.18 bead/g of soil sample to 587 beads/g of soil, while soil samples taken near unmarked roads had average range of concentration of 0 bead/g of soil to 0.21 bead/g of soil. Retroreflective glass beads come from pavement markings, thus soil samples near marked roads are expected to have higher concentrations of glass beads. Analysis of spatial distribution of glass beads showed that as sample collection moved further from the road, concentration of glass beads decreased.
ICP-MS results of elemental concentrations for each sample showed discriminative differences between samples, for most of the elements. An analysis of variance for elemental concentrations was conducted, and results showed statistically significant differences, beyond random chance alone for half of the elements analyzed. For forensic comparisons, a significant difference in even just one element is enough to conclude that the samples came from different sources. The elemental concentrations of glass beads collected from the same location, but of varying differences, was also analyzed. ANOVA results show significant differences for only one or two elements. A pair-wise t-test was conducted to determine which elements are most discriminative among all the samples. Rubidium was found to be the most discriminative, showing significant difference for 67% of the pairs. Beryllium, potassium, and manganese were also highly discriminative, showing significant difference for at least 50% of all the pairs.
ContributorsGomez, Janelle Kate Pacifico (Author) / Montero, Shirly (Thesis advisor) / Herckes, Pierre (Thesis advisor) / Borges, Chad (Committee member) / Gordon, Gwyneth (Committee member) / Arizona State University (Publisher)
Created2023
Description
Linear-regression estimators have become widely accepted as a reliable statistical tool in predicting outcomes. Because linear regression is a long-established procedure, the properties of linear-regression estimators are well understood and can be trained very quickly. Many estimators exist for modeling linear relationships, each having ideal conditions for optimal performance. The differences stem from the introduction of a bias into the parameter estimation through the use of various regularization strategies. One of the more popular ones is ridge regression which uses ℓ2-penalization of the parameter vector. In this work, the proposed graph regularized linear estimator is pitted against the popular ridge regression when the parameter vector is known to be dense. When additional knowledge that parameters are smooth with respect to a graph is available, it can be used to improve the parameter estimates. To achieve this goal an additional smoothing penalty is introduced into the traditional loss function of ridge regression. The mean squared error(m.s.e) is used as a performance metric and the analysis is presented for fixed design matrices having a unit covariance matrix. The specific problem setup enables us to study the theoretical conditions where the graph regularized estimator out-performs the ridge estimator. The eigenvectors of the laplacian matrix indicating the graph of connections between the various dimensions of the parameter vector form an integral part of the analysis. Experiments have been conducted on simulated data to compare the performance of the two estimators for laplacian matrices of several types of graphs – complete, star, line and 4-regular. The experimental results indicate that the theory can possibly be extended to more general settings taking smoothness, a concept defined in this work, into consideration.
ContributorsSajja, Akarshan (Author) / Dasarathy, Gautam (Thesis advisor) / Berisha, Visar (Committee member) / Yang, Yingzhen (Committee member) / Arizona State University (Publisher)
Created2022
Description
Type 1 diabetes (T1D) is the result of an autoimmune attack against the insulin-producing β-cells of the pancreas causing hyperglycemia and requiring the individual to rely on life-long exogenous insulin. With the age of onset typically occurring in childhood, there is increased physical and emotional stress to the child as well as caregivers to maintain appropriate glucose levels. The majority of T1D patients have antibodies to one or more antigens: insulin, IA-2, GAD65, and ZnT8. Although antibodies are detectable years before symptoms occur, the initiating factors and mechanisms of progression towards β-cell destruction are still not known. The search for new autoantibodies to elucidate the autoimmune process in diabetes has been slow, with proteome level screenings on native proteins only finding a few minor antigens. Post-translational modifications (PTM)—chemical changes that occur to the protein after translation is complete—are an unexplored way a self-protein could become immunogenic. This dissertation presents the first large sale screening of autoantibodies in T1D to nitrated proteins. The Contra Capture Protein Array (CCPA) allowed for fresh expression of hundreds of proteins that were captured on a secondary slide by tag-specific ligand and subsequent modification with peroxynitrite. The IgG and IgM humoral response of 48 newly diagnosed T1D subjects and 48 age-matched controls were screened against 1632 proteins highly or specifically expressed in pancreatic cells. Top targets at 95% specificity were confirmed with the same serum samples using rapid antigenic protein in situ display enzyme-linked immunosorbent assay (RAPID ELISA) a modified sandwich ELISA employing the same cell-free expression as the CCPA. For validation, 8 IgG and 5 IgM targets were evaluated with an independent serum sample set of 94 T1D subjects and 94 controls. The two best candidates at 90% specificity were estrogen receptor 1 (ESR1) and phosphatidylinositol 4-kinase type 2 beta (PI4K2B) which had sensitivities of 22% (p=.014) and 25% (p=.045), respectively. Receiver operating characteristic (ROC) analyses found an area under curve (AUC) of 0.6 for ESR1 and 0.58 for PI4K2B. These studies demonstrate the ability and value for high-throughput autoantibody screening to modified antigens and the frequency of Type 1 diabetes.
ContributorsHesterman, Jennifer (Author) / LaBaer, Joshua (Thesis advisor) / Borges, Chad (Committee member) / Sweazea, Karen (Committee member) / Mangone, Marco (Committee member) / Arizona State University (Publisher)
Created2022
Description
Parkinson’s Disease is one of the most complicated and abundantneurodegenerative diseases in the world. Previous analysis of Parkinson’s disease has
identified both speech and gait deficits throughout progression of the disease. There has
been minimal research looking into the correlation between both the speech and gait
deficits in those diagnosed with Parkinson’s. There is high indication that there is a
correlation between the two given the similar pathology and origins of both deficits. This
exploratory study aims to establish correlation between both the gait and speech deficits
in those diagnosed with Parkinson’s disease. Using previously identified motor and
speech measurements and tasks, I conducted a correlational study of individuals with
Parkinson’s disease at baseline. There were correlations between multiple speech and gait
variability outcomes. The expected correlations ranged from average harmonics-to-noise
ratio values against anticipatory postural adjustments-lateral peak distance to average
shimmer values against anticipatory postural adjustments-lateral peak distance. There
were also unexpected outcomes that ranged from F2 variability against the average
number of steps in a turn to intensity variability against step duration variability. I also
analyzed the speech changes over 1 year as a secondary outcome of the study. Finally, I
found that averages and variabilities increased over 1 year regarding speech primary
outcomes. This study serves as a basis for further treatment that may be able to
simultaneously treat both speech and gait deficits in those diagnosed with Parkinson’s.
The exploratory study also indicates multiple targets for further investigation to better
understand cohesive and compensatory mechanisms.
ContributorsBelnavis, Alexander Salvador (Author) / Peterson, Daniel (Thesis advisor) / Daliri, Ayoub (Committee member) / Berisha, Visar (Committee member) / Arizona State University (Publisher)
Created2022
Description
In the era of big data, more and more decisions and recommendations are being made by machine learning (ML) systems and algorithms. Despite their many successes, there have been notable deficiencies in the robustness, rigor, and reliability of these ML systems, which have had detrimental societal impacts. In the next generation of ML, these significant challenges must be addressed through careful algorithmic design, and it is crucial that practitioners and meta-algorithms have the necessary tools to construct ML models that align with human values and interests. In an effort to help address these problems, this dissertation studies a tunable loss function called α-loss for the ML setting of classification. The alpha-loss is a hyperparameterized loss function originating from information theory that continuously interpolates between the exponential (alpha = 1/2), log (alpha = 1), and 0-1 (alpha = infinity) losses, hence providing a holistic perspective of several classical loss functions in ML. Furthermore, the alpha-loss exhibits unique operating characteristics depending on the value (and different regimes) of alpha; notably, for alpha > 1, alpha-loss robustly trains models when noisy training data is present. Thus, the alpha-loss can provide robustness to ML systems for classification tasks, and this has bearing in many applications, e.g., social media, finance, academia, and medicine; indeed, results are presented where alpha-loss produces more robust logistic regression models for COVID-19 survey data with gains over state of the art algorithmic approaches.
ContributorsSypherd, Tyler (Author) / Sankar, Lalitha (Thesis advisor) / Berisha, Visar (Committee member) / Dasarathy, Gautam (Committee member) / Kosut, Oliver (Committee member) / Arizona State University (Publisher)
Created2022
Description
Plasticizers are plastic additives used to enhance the physical properties of plastic and are ubiquitous in the environment. A class of plasticizer compounds called phthalate esters that are not fully eliminated in wastewater treatment facilities are relevant to the ecological health of downstream ecosystems and urban areas due to their ecotoxicity, tendency for soil accumulation, and the emerging concern about their effects on public health. However, plasticizer concentrations in a constructed wetland environment have rarely been studied in the United States, prompting the need for a method of plasticizer quantification in the Tres Rios Constructed Wetlands which are sustained by the effluent of the 91st Avenue Wastewater Treatment Plant in Phoenix, Arizona. The concentrations of four common plasticizer compounds (dimethyl: DMP, diethyl: DEP, di-n-butyl: DnBP, and bis(2-ethylhexyl): DEHP phthalate) at five sites across the wetland surface water were quantified using solid-phase extraction followed by gas chromatography coupled with mass spectrometry (GC/MS). The sampling period included four sample sets taken from March 2022 to September 2022, which gave temporal data in addition to spatial concentration data. Quantification and quality control were performed using internal standard calibration, replicate samples, and laboratory blanks. Higher molecular weight phthalates accumulated in the wetland surface water at significantly higher average concentrations than those of lower molecular weight at a 95% confidence level, ranging from 8 ng/L to 7349 ng/L and 4 ng/L to 27876 ng/L for DnBP and DEHP, respectively. Concentrations for dimethyl phthalate and diethyl phthalate were typically less than 50 ng/L and were often below the method detection limit. Average concentrations of DnBP and DEHP were significantly higher during periods of high temperatures and arid conditions. The spatial distribution of phthalates was analyzed. Most importantly, a method for successful ultra-trace quantification of plasticizers at Tres Rios was established. These results confirm the presence of plasticizers at Tres Rios and a significant seasonal increase in their surface water concentrations. The developed analytical procedure provides a solid foundation for the Wetlands Environmental Ecology Lab at ASU to further investigate plasticizers and contaminants of emerging concern and determine their ultimate fate through volatilization, sorption, photodegradation, hydrolysis, microbial biodegradation, and phytoremediation studies.
ContributorsStorey, Garrett (Author) / Herckes, Pierre (Thesis director) / Childers, Dan (Committee member) / Borges, Chad (Committee member) / Barrett, The Honors College (Contributor) / School of Sustainability (Contributor) / School of Molecular Sciences (Contributor)
Created2023-05
Description
There are limited methods and techniques to quantitatively assess protein content in single cells or small cell populations of tissues. The standard protein insulin was used to understand how potential changes in the preparation or co-crystallization process could improve sensitivity and limit of detection through matrix assisted laser desorption ionization (MALDI) mass spectrometry analysis in Bruker’s Microflex LRF using polydimethylsiloxane (PDMS) reservoirs. In addition, initial imaging tests were performed on Bruker’s RapifleX MALDI Tissuetyper to determine the instrument’s imaging capabilities on proteins of interest through the use of a single layer “Christmas tree” microfluidic device, with the aim of applying a similar approach to future tissue samples. Data on 2µM insulin determined that a 95% laser power in the Microflex corresponded to 12-15% laser power in the RapifleX. Based on the experiments with insulin, the process of mixing insulin and saturated ɑ-Cyano-4-hydroxycinnamic acid (HCCA) matrix solvent in a 1:1 ratio using 10mM sodium phosphate buffer under area analysis is most optimized with a limit of detection value of 110 nM. With this information, the future aim is to apply this method to a double layer Christmas tree device in order to hopefully quantitatively analyze and image protein content in single or small cell populations.
ContributorsKow, Keegan (Author) / Ros, Alexandra (Thesis director) / Borges, Chad (Committee member) / Cruz-Villarreal, Jorvani (Committee member) / Barrett, The Honors College (Contributor) / School of Molecular Sciences (Contributor)
Created2023-05
Description
Antibodies are the immunoglobulins which are secreted by the B cells after a microbial invasion. They are stable and stays in the serum for a long time which makes them an excellent biomarker for disease diagnosis. Inflammatory bowel disease is a type of autoimmune disease where the immune system mistakenly attacks the commensal bacteria and leads to inflammation. We studied antibody response of 100 Crohn’s disease (CD), 100 ulcerative colitis (UC) and 100 healthy controls against 1,173 bacterial and 397 viral proteins. We found some anti-bacterial antibodies higher in CD compared to controls while some antibodies lower in UC compared to controls. We were able to build biomarker panels with AUCs of 0.81, 0.87, and 0.82 distinguishing CD vs. control, UC vs. control, and CD vs. UC, respectively. Subgroup analysis based on the Montreal classification revealed that penetrating CD behavior (B3), colonic CD location (L2), and extensive UC (E3) exhibited highest antibody reactivity among all patients. We also wanted to study the reason for the presence of autoantibodies in the sera of healthy individuals. A meta-analysis of 9 independent biomarker study was performed to find 77 common autoantibodies shared by healthy individuals. There was no gender bias; however, the number of autoantibodies increased with age, plateauing around adolescence. Molecular mimicry likely contributed to the elicitation of a subset of these common autoantibodies as 21 common autoantigens had 7 or more ungapped amino acid matches with viral proteins. Intrinsic properties of protein like hydrophilicity, basicity, aromaticity, and flexibility were enriched for common autoantigens. Subcellular localization and tissue expression analysis indicated the sequestration of some autoantigens from circulating autoantibodies can explain the absence of autoimmunity in these healthy individuals.
ContributorsShome, Mahasish (Author) / LaBaer, Joshua (Thesis advisor) / Borges, Chad (Committee member) / Stephanopoulos, Nicholas (Committee member) / Arizona State University (Publisher)
Created2021