UMBC Theses and Dissertations

Permanent URI for this collectionhttp://hdl.handle.net/11603/63

Browse

Now showing 1 - 20 of 1391

A Cross-national Comparison of COVID-19 and Mental Health: Understanding Social Trust.
(2024/01/01) Hunter-Cevera, Jonathan; McLaren, Zoe ZM Adler, Marina MA; School of Public Policy; Public Policy
The COVID-19 pandemic caused a global challenge that many countries were not well equipped for. The present study examined why certain countries did better in terms of deaths caused from the novel virus and mental health outcomes during the first 17 months of the pandemic. Current research has examined many government and policy related variables at an attempt to explain differences of COVID-19 outcomes between and within countries. Studies also examine the complicated changing concept of social capital as well. The goal of this research was to examine countries (N=50) in a cross-national setting testing government/economic, welfare states, Non-pharmaceutical intervention policies, and social trust variables on deaths from COVID-19 and adverse mental health. Various multiple regression series were run using mediation, moderation, and quadratic techniques. The results suggested that having higher social trust in the form of trusting other citizens and trusting government and public institutions may have had a strong influence in reducing countries COVID-19 deaths and adverse mental health outcomes. Countries’ public health officials and government should take into account the public levels of social trust before implementing NPIs and vaccination rollout. A country’s level of social trust may be an important factor to successfully navigate through a pandemic.
Bouncing Forward: Exploring Nonprofit Resilience Through Emergency Management
(2024/01/01) De Oro, Grace Catherine; Hamilton Edwards, Lauren; School of Public Policy; Public Policy
Resilience is often associated with the ability to recover quickly or bounce back. However, often, recovery does not look like a return to "normal" or homeostasis. Nonprofit organizations are essential to providing services in response to emergencies, and their involvement builds communities' capacity to deal with future disasters. Nonprofit organizations' response to the Coronavirus (COVID-19) pandemic has grown almost as quickly as the news and fear of the virus spreading across all continents. The pandemic challenged nonprofits' limits, management, and preparedness everywhere, often exposing limited capacity to bounce back while helping the populations and communities they serve. This dissertation research focuses on two major exploratory questions by utilizing a systematic literature review and a case study on emergency management nonprofits. First, what is resilience, and how can we measure it? and second, what does organizational resilience look like in nonprofit organizations? The systematic literature review and theoretical background of interdisciplinary literature establish this study of resilience from the perspective of nonprofit organizations. The literature is used to develop a framework in which resilience is a process, not a characteristic of an individual, organization, or community. By conducting a case study on emergency management nonprofits, this dissertation examines how the experience and response to the COVID-19 pandemic impacted the organization through primary data collection and interviews with nonprofit practitioners. Risk avoidance, risk management, and the control of risk, with the capacity to cope resiliently, are used to build a relationship between resilience and emergency management (Wildavsky,1998). The findings of this study have direct implications for how future scholarship defines resilience and stresses the importance of actively including and involving practitioners in the field. The COVID-19 pandemic provides a unique situation as risk and perceived risk were completely unknown. Nonprofit organizations acted in the best way they knew how and to their capacity. The findings also include a conversation on mission pivot, the short-term and long-term impacts of a disaster like COVID-19, and participants' reactions to COVID-19.
The Roles of Chinese American Mothers’ Stress and Parenting in their Children’s Socio-Emotional and Behavioral Outcomes
(2024/01/01) Wang, Suqing; Cheah, Charissa C. S. L. C; Psychology; Psychology
Parental factors play crucial roles in children’s socio-emotional and behavioral (SEB) outcomes. Despite being the largest subgroup within a rapidly expanding Asian American community in the United States, these processes remain understudied in Chinese American families, particularly given the cultural variations in parenting practices. Across three papers, this dissertation project aimed to explore the determinants of children’s SEB outcomes and the roles of parents within the context of Chinese American families. The first paper examined the associations between Chinese American mothers’ experiences of two types of stress (i.e., general/contextual stress and parenting stress) and their children’s SEB adjustment problems as well as the mediating roles of maternal psychologically controlling parenting (PCP) and maternal warmth in these associations. The results revealed that higher levels of general/contextual stress and parenting stress were each uniquely associated with more maternal PCP practices, which in turn was associated with fewer child adjustment problems. Although surprising, the finding that maternal PCP did not negatively impact child adjustment aligned with some previous research that suggests that such parenting practices are perceived less negatively in interdependence-oriented cultures, such as Chinese culture. These findings highlighted the critical role of cultural context in interpreting the impacts of parenting practices on child adjustment. The second and third papers focused on specific parenting practices, maternal emotion-related socialization behaviors. The second paper explored the mediating role of children’s effortful control in the associations between Chinese American mothers’ positive and negative emotional expressiveness and their children’s SEB adjustment problems. Results indicated that higher maternal positive emotional expressiveness and lower negative emotional expressiveness were each uniquely associated with higher child effortful control, which in turn was associated with fewer child adjustment problems. Our results demonstrated that the positive emotional support provided by Chinese American mothers to children played a crucial role in shaping their children's effortful control, a fundamental capacity that enhances child SEB functioning. The third paper utilized an observational task that is designed to evoke negative emotions in children to examine the associations between maternal responses to child emotions (physical soothing and symbolic/cognitive soothing) and children’s expressed positivity and sadness, as well as the moderating roles of general/contextual stress and parenting stress. The findings revealed that mothers’ parenting stress moderated the relations between mothers’ soothing behaviors and children’s expressed positivity such that children expressed higher positivity following physical soothing only when mothers perceived experiencing low levels of parenting stress. Conversely, children showed lower positivity following physical soothing if mothers perceived experiencing high levels of parenting stress. These contrasting moderation effects highlighted the importance of considering maternal parenting stress in emotion socialization processes. Together, this dissertation project expanded the literature, especially in the areas of emotion socialization and stress theories, by highlighting specificity in processes and the importance of culturally informed research that can inform interventions to improve the development and adjustment of Chinese American families.
The Influence of Psychiatric Hospitalization on Mental Health Help-Seeking for LGBTQ People of Color
(2024/01/01) Jumarali, Selima; Nnawulezi, Nkiru; Psychology; Psychology
Lesbian, gay, bisexual, transgender, and queer (LGBTQ) people of color (POC) may be uniquely impacted by psychiatric hospitalization and at greater risk of reporting perceived coercion, which is associated with increase suicide risk. This study examined the prevalence of perceived coercion in a sample of 133 LGBTQ POC who were psychiatrically hospitalized for suicidality and investigated whether perceived coercion predicts traumatic impact and future help-seeking intent using hierarchical linear regression models. Seventy-one percent of the sample reported perceived coercion and 29% of the sample reported no perceived coercion. Transgender and gender expansive (TGE) participants reported higher levels of perceived coercion than men. Perceived coercion independently significantly predicted traumatic impact and future help-seeking intent. Needs satisfaction and institutional betrayal each significantly predicted traumatic impact and future help-seeking intent when controlling for perceived coercion. Twenty-seven percent of participants encountered police involvement on their journey to hospitalization, 74% of whom reported a negative experience with the police. Clients who experienced police involvement were more likely to report lower needs satisfaction, greater institutional betrayal, greater traumatic impact, and lower future help-seeking intent. Medical and mental health providers, as well as administrators and staff, in inpatient psychiatric units can promote needs satisfaction through prioritizing patient autonomy, affirming patients’ strengths, demonstrating care and friendliness toward all patients, and cultivating social connection and peer support among patients. Institutions can prevent betrayal through LGBTQ+ affirming care practices and cultural competency training to avoid incidents of bias and discrimination. Prevention also includes publicizing patients’ rights, posting guidelines for behavior to promote an inclusive climate, and having highly visible methods to report bias and dissatisfaction with care. Institutions can also work to repair betrayal by incentivizing providers to solicit patient input in treatment, listen to patient feedback with empathy, apologize for missteps, and share additional steps patients can take for the institutions to address their grievances. Future research should investigate alternatives to hospitalization including non-police models of crisis response. Keywords: Inpatient Hospitalization, Perceived Coercion, LGBTQ, People of Color, Help-Seeking, Traumatic Impact, Institutional Betrayal, Needs Satisfaction, Police
The Impact of Observation and Distraction on Pediatric Acute Pain
(2024/01/01) Jehl, Brianna Lynn; Dahlquist, Lynnda; Psychology; Psychology
The present study was a novel exploration of the impacts of observation and distraction on acute pain outcomes in elementary school aged children. A sample of 96 children (40 girls) between the ages of 6 and 13 underwent two cold pressor trials during which pain tolerance, nervousness, pain intensity, and pain unpleasantness were assessed. The first trial served as a baseline; no intervention was conducted. During the second trial, participants were assigned to one of four experimental conditions—control (a repeat of baseline), observer present (participants completed the trial in the presence of an observer they were told was monitoring how well they performed), VR distraction (participants played Beat Saber while completing the trial) or observer present + VR (participants played the VR game while also being observed). It was hypothesized that the presence of an observer would decrease cold pressor tolerance and increase nervousness, pain intensity, and pain unpleasantness. It was also hypothesized that the use of distraction would increase cold pressor tolerance and decrease nervousness, pain intensity, and pain unpleasantness. The presence of the observer was expected to increase participant stress and result in poorer pain outcomes, whereas VR distraction was expected to result in improved pain outcomes and potentially counteract the negative impact of observation. The results revealed the expected effect of VR distraction on pain tolerance. Children who utilized VR had greater increases in pain tolerance than those children who did not utilize VR (p < .001). However, the presence of the observer did not significantly impact pain tolerance and there was not a significant interaction between observer presence and VR distraction. The expected interaction between observer presence and VR distraction was only obtained for unpleasantness. Pain unpleasantness significantly increased for those children in the observer present condition during trial 2 compared to trial 1 (p = .028) and decreased for those children in the observer present + VR condition during trial 2 (p = .025). However, unpleasantness did not decrease for those children in the VR only condition, making these findings difficult to interpret. A main effect of trial was found for nervousness such that children were more nervous during the second trial compared to the first (p = .005). Pain intensity was impacted by gender such that girls, not boys, who did not receive VR distraction experienced a significant decrease in pain intensity during trial 2 compared to trial 1 (p = .019), which was the opposite of what was expected. While observation was largely insignificant in the current study, it did have an impact on children’s pain unpleasantness; therefore, it is important to continue studying observation. It is likely that the observer was not potent enough to have a true impact on pain outcomes in the current study. Further research efforts should utilize diverse samples, multiple measurements of stress, and qualitative measures in order to allow for a more nuanced understanding of the impact of observation and distraction on pain.
A Multi-Method Examination of Asian American Parents and Adolescents’ Ethnic-Racial and Civic Socialization
(2024/01/01) Cho, Hyun Su; Cheah, Charissa S. L.; Psychology; Psychology
During the COVID-19 outbreak, the racialization of the pandemic refueled and intensified interpersonal and collective racial prejudice against Asian Americans, adversely impacting these families’ adjustment in both shorter and longer terms. In addition, the police murder of George Floyd in 2020 ignited an increase in momentum for the Black Lives Matter (BLM) movement, which led many Asian Americans to reflect on systemic racism and their social positioning. Amidst this critical racial reckoning, ethnic-racial and civic socialization within the family context can shape the way children and adolescents view themselves as minoritized individuals and how they navigate the racialized landscape and address social inequity issues in the United States. Therefore, across three papers, this dissertation project broadly aimed to understand how Asian American families’ ethnic-racial and civic socialization is shaped and associated with adolescents’ civic engagement. Briefly, the first paper examined the association between Chinese American parents’ racial discrimination experiences and their maintenance of their heritage culture and concealing their Chinese heritage and connection ethnic-racial socialization during the COVID-19 pandemic, the mediating role of parents’ psychological well-being, the moderating role of family support, and the potential variations in these associations among parents with children across three developmental stages. The second paper focused on the mediating role of Chinese American parents’ civic socialization in the association between their interpersonal racial discrimination experiences and their adolescent’ community and political engagement. The potential exacerbating role of parents’ concerns about structural racism was explored. Finally, the third paper explored Asian American adolescents’ ethnic-racial and civic beliefs and responses to anti-Asian hate and the BLM movement during discussions with their parents. Moreover, within-group patterns among Chinese, Filipino, and Korean American adolescents’ responses, reasons, and their associations with their critical consciousness were examined. Together, the three studies made important empirical and theoretical contributions to the field by revealing: (1) underlying mechanisms in the association between racial discrimination and Asian American parents’ ethnic-racial and civic socialization and their adolescents’ civic engagement; (2) Asian American families’ resilience in the context of the rise in anti-Asian hate and racial tension; and (3) Asian American adolescents’ contributions to parent-adolescent discussions about race and social change. The present dissertation project has significant practical implications and calls for efforts to address systemic racism to ensure conscious-raising systems that can facilitate the positive development and thriving of Asian American children, adolescents, and families.
Reasons Why: Contributors of Suicide Risk among a Community Sample of Young People with Psychosis-like Experiences
(2024/01/01) Jay Title, Samantha; Schiffman, Jason; Pitts, Steven; Psychology; Psychology
Adolescents and young adults who experience higher levels of psychosis-like experiences (PLEs) also experience elevated rates of suicidal thoughts, suicide attempts, and deaths by suicide, as compared to the general population. Though there has been a recent burst of literature examining the relation between PLEs and suicidal thoughts and behaviors (STB), there is still a limited understanding of the factors that contribute to the heightened risk for STB among people with higher levels of PLEs. Among a sample of young people with higher levels of PLEs (N = 234), this study introduced The Suicide History Assessment for People with Psychosis-spectrum Experiences (SHAPE), a suicide risk assessment developed to measure relevant general and psychosis-spectrum-specific risk factors of STB. The present study (1) evaluated the relation between subtypes of PLEs and STB, (2) compared clinical characteristics of youth who reported that their PLEs impacted their suicidal thoughts to those who did not report a connection, (3) identified participant-rated general risk factors of suicidal thoughts among youth who experienced higher levels of PLEs, and (4) examined the sum of participant-identified general risk factors as a mediator of the relation between PLEs and STB. Hypotheses were partially supported. Results from this pilot study confirmed a relation between PLEs and suicide risk severity, as well as differential relations between PLE subtypes and suicide risk severity. A subset of participants reported that subtypes of PLEs impacted their suicidal thoughts, and this subsample presented with more clinically severe symptoms than those who did not identify a connection between symptoms. General suicide risk factors contributed to suicidal thoughts among people with high levels of PLEs, and these risk factors differed based on whether participants experienced high versus low levels of PLEs. Finally, the sum of these risk factors mediated the relation between distressing PLEs and suicide risk severity. Information gleaned from this study informed revisions and additions to the SHAPE and helped to identify relevant risk factors for STB among people with PLEs.
Relations among Attenuated Psychosis Symptoms, Internalized Stigma, and Help-Seeking Intentions in a Community Sample of Youth
(2024/01/01) Akouri-Shan, LeeAnn; Pitts, Steven C Schiffman, Jason; Psychology; Psychology
Attenuated psychosis symptoms often emerge during adolescence and young adulthood and, for some youth may be distressing, impairing, and/or increase in severity over time. Early identification of these youth combined with early intervention efforts can prevent or mitigate a host of negative outcomes traditionally associated with full-threshold psychotic disorders. Internalized mental health stigma has been widely established as a significant barrier to help-seeking among youth with mental health concerns, particularly those with attenuated psychosis symptoms. In addition to stigma, various symptom-related factors (e.g., symptom type, severity, and related distress) can also influence help-seeking among this group. Although positive symptom severity is thought to be positively associated with help-seeking among youth in the general population, relations between other types of symptoms – e.g., negative symptoms or affective symptoms – and help-seeking remain somewhat unclear. The current study examined relations among self-reported mental health symptoms, internalized stigma, and intent to seek mental health treatment in a community sample of youth who were not currently engaged in treatment (N = 601, M age = 21.24). Participants were categorized into two clinical groups – those who scored at or above threshold, and those who scored below threshold on psychosis risk screening tools. Internalized stigma partially mediated the relation between positive symptom severity and intent to seek treatment and was positively associated with intent to seek treatment in the sample. Negative symptoms and general symptoms were each positively associated with internalized stigma and intent, controlling for positive symptoms. The relation between internalized stigma and intent did not appear to differ by clinical group. Among different types of internalized stigma, alienation was uniquely associated with intent to seek treatment in the full sample, with higher levels of alienation predicting greater intent. Findings highlight the complex interplay between symptoms, stigma, and help-seeking experiences among community youth and may inform early intervention efforts for this population.
Oppression and Anti-oppression in Clinical Psychology
(2023-01-01) Abraham, Munazza Saalim; Hunter, Bronwyn; Murphy, Chris; Psychology; Psychology
Traditional clinical psychology programs do not require anti-oppression training or specific social justice and empowerment practices, compared to some social work and counseling programs. Therefore, it follows that clinical psychology students, faculty, and practitioners rarely know how or when to address oppression with diverse groups and within diverse contexts. The current pilot study represents initial steps towards developing anti-oppression training for clinical psychologists and trainees, grounded in the bystander intervention theory by Latané and Darley (1970). The first major step was to conduct in-depth interviews to qualitatively explore how clinical psychologists and trainees’ have witnessed, experienced, and/or perpetuated oppression in clinical psychology settings (classroom, supervision, and therapy settings); and, based on their experiences, what are suggestions towards anti-oppression. This first step informs anti-oppression training goals and content. The second major step was to survey a diverse range of clinical psychologists and trainees from across the United States — such as graduate student trainees, faculty members, and practicing clinicians — to also quantitatively assess to what extent they have witnessed, experienced, and/or perpetuated oppression in their field as well as assess the acceptability and feasibility of anti-oppression training in clinical psychology. This step further informs anti-oppression training goals as well as structure. This research is important, because, while the literature may support the need for anti-oppression training, the prospective trainees and training programs must also acknowledge this need to initiate training implementation and success. Results showed that among clinical psychologists and trainees in the sample (N = 51), the majority reported witnessing and personally experiencing oppression, and a little less than a quarter reported perpetuating oppression in clinical psychology settings. Additionally, the majority of participants reported high perceived responsibility, yet relatively low confidence in their current skills to address oppression. Furthermore, participants reported substantial acceptability and perceived need for anti-oppression training while also providing several suggestions on what anti-oppression training can look like (i.e., decentering whiteness, learning how to identify oppression, role playing on how to address oppression, etc.). While additional research may be needed, study results can inform future training for clinical psychologists and trainees to become active bystanders in anti-oppression praxis.
Linear Computing and Thermodynamics
(2024/01/01) Aifer, Maxwell; Deffner, Sebastian; Physics; Physics
A large part of computing deals with the doing, undoing, and reinterpretation of linear transformations on vector spaces. These operations are essential to the description of most physical phenomena, are a core part of machine learning, and also form the basic building blocks of quantum algorithms. The tendency of all things to evolve towards thermal equilibrium poses challenges for efficient and accurate computation, which often appear in the form of energy dissipation and errors caused by thermal noise. While thermal physics has provided some insights about physical limitations on computing, less is known about the particular considerations that apply to linear algebraic computations. Here, I study the application of thermodynamics to such problems, in both classical and quantum systems. Several new algorithms are given for linear algebra problems using specialized hardware, which have theoretical advantages over previously existing algorithms. Additionally, new bounds are proven for the energy cost of unitary transformations of quantum states, as well as for the efficiency of communication between quantum systems.
ANALYSIS OF SMYD1 FUNCTION AND THE MOLECULAR MECHANISM OF ACTION IN SKELETAL AND CARDIAC MUSCLE CELL DIFFENCIATION
(2024/01/01) Xu, Rui; Du, Shaojun; Zohar, Yonathan; Marine-Estuarine Environmental Sciences; Marine-Estuarine-Environmental Sciences
Smyd1, a key lysine methyltransferase, is crucial for myofibrillogenesis in both skeletal and cardiac muscles. The zebrafish orthologue, smyd1b, encodes two isoforms, Smyd1b_tv1 and Smyd1b_tv2, differing by 13 amino acids due to alternative splicing. It remains unclear whether the Smyd1b_tv1 and Smyd1b_tv2 isoforms exhibit distinct expression patterns and functions. The Janus face of Smyd1 in muscle cell development is perplexing and controversial. Some lines of evidence have suggested that Smyd1 functions as a chromatin modifier to regulate gene expression through a common post-translational mechanism. However, other lines of evidence indicate that Smyd1 may act as a chaperone protein in the cytosol to stabilize sarcomeric-associated proteins. The factors that trigger the enhanced protein degradation in muscles upon loss of Smyd1 function remain unknown. To determine whether Smyd1b_tv1 and Smyd1b_tv2 may have distinct function in muscle development, we first generated transgenic zebrafish models that express only Smyd1b_tv1 or Smyd1b_tv2 isoform in a zebrafish mutant that lacks the endogenous smyd1b gene. Functional analysis revealed that expression of Smyd1b_tv1 is essential for cardiomyocyte differentiation and fish viability, while Smyd1b_tv2 is dispensable for heart development and fish survival. Targeted deletion of Smyd1b_tv1 using CRISPR/Cas9 to produce smyd1b mutants that only express endogenous smyd1b_tv2, referred to as Smyd1bTV2. Surprisingly, Smyd1bTV2 fish were viable, and showed no apparent phenotypes. However, upon on the cardiac injury in adult fish, Smyd1bTV2 fish showed poor heart regeneration compared to WT siblings, emphasizing the critical role of Smyd1b_tv1 in cardiac regeneration. To elucidate the subcellular functionality of Smyd1b in orchestrating muscle development in zebrafish, multiple zebrafish transgenic lines expressing nuclear (Smyd1bNLS) and cytosolic (Smyd1bNES) Smyd1b were established. We showed that cytosolic Smyd1b (Smyd1bNES) completely complements Smyd1b-deficient mutants, while nuclear Smyd1b (Smyd1bNLS) exacerbates heart defects, highlighting the importance of Smyd1b cytosolic localization. Furthermore, comprehensive RNA-seq analysis revealed that smyd1b exhibits strikingly similar transcriptional profiles to unc45b, a chaperone mutant. In smyd1b-/- mutants, misfolded Myosin forms aggregate positive for ubiquitin during myogenesis, indicating Smyd1b primarily acts as a chaperone for Myosin folding, and the ubiquitin-proteasome system (UPS) may play an important role in deconstructing misfolded Myosin. Further investigation into specific UPS components, including CHIP and UBA1, showed that their loss did not ameliorate the sarcomeric defects or prevent sarcomeric protein from degradation in smyd1b-/- mutants. These findings imply the involvement of multiple E3 ligases in this process and the complexity of the protein quality control mechanisms in maintaining sarcomere integrity. Overall, Smyd1b primarily functions as a chaperone in the cytoplasm, aiding in myosin folding and sarcomere assembly. The degradation of misfolded myosin in smyd1b-/- mutants may be related to the ubiquitin-proteasome system (UPS). The two isoforms encoded by Smyd1b are not functionally equivalent, with Smyd1b_tv1 playing a more critical role in cardiac muscle development and regeneration.
Semigroup Methods for Poroelastic Multi-Physics Systems Describing Biological Tissues
(2024/01/01) Gurvich, Elena; Webster, Justin T; Mathematics and Statistics; Mathematics, Applied
This thesis presents novel work in the mathematical theory of poroelasticity, which was first phenomenologically developed by Biot and Zenisek during the mid-last century. The theory relates saturated porous structural deformations to fluid pressure changes within, and blossomed through many applications in the geosciences (e.g., seismic and petroleum engineering). At the outset of the 21st century, poroelasticity proved to be a revolutionary incorporation to the biological fields (e.g. biomedical engineering, arterial stents, scaffolding), owing to the poroelastic nature of biological tissues. For the parameters of physical interest, a quasi-static approximation induces dynamics, which can be represented as an implicit evolution. Moreover, compressibility in Biot's equations is a significant consideration. In the incompressible limit, Biot’s model degenerates This dissertation will present a biologically motivated multilayered system, composed of the coupled dynamics of a 3D poroelastic structure, a poroelastic plate, and an incompressible free Stokes flow. We propose two constituent sub-problems, to gain a better understanding of this extremely complex system. First, a complete well-posedness analysis of the poroelastic plate is shown utilizing variational tools. Secondly, Biot-Stokes filtration is proposed with Beavers-Joseph-Saffman coupling conditions on a fixed 2D interface. A semigroup approach is used to bypass the issues with mismatched trace regularities on the interface; thus guaranteeing strong and generalized solutions. Then the existence of weak solutions, including the degenerate case, is provided by argument by density. The most interesting cases are singular limits, which lead to the use of the theory of abstract implicit, degenerate evolutions, of which the appendix supplies a brief overview. Thus, this thesis provides a clear elucidation of strong solutions and the construction of weak solutions for inertial linear Biot-Stokes filtration systems and uniquely for a poroelastic plate, as well as their regularity through associated estimates.
Fair and Interpretable Pseudo Value-Based Deep Learning Models for Federated Survival Analysis
(2024/01/01) Rahman, Md Mahmudur; Purushotham, Sanjay; Information Systems; Information Systems
Survival analysis, or time-to-event analysis, aims to predict the time until an event occurs, providing valuable insights into the temporal aspects of various phenomena, such as disease progression. This dissertation addresses the growing need for fair and interpretable machine learning models in survival analysis within the healthcare domain, alongside the necessity for privacy-preserving distributed training methods to enhance generalization and data utilization. In particular, we focus on the following problems: 1) how to efficiently handle censoring, i.e., incomplete survival outcomes; 2) how to make unbiased estimations of the survival analysis quantities in the presence of competing risks and multi-state transitions; 3) how to enhance the interpretability and fairness of survival analysis models; and 4) how to enable privacy-preserving distributed training of survival models to address limited data utilization and lack of generalization due to strict privacy laws such as GDPR and HIPAA. In this dissertation, we provide the following solutions to these problems: 1) To address the censoring challenge, we utilize theoretically consistent pseudo-values from statistical paradigms, simplifying the problem to a regression analysis task. 2) To provide unbiased survival predictions for complex problems like competing risk and multi-state survival analysis, we introduce novel pseudo-value-based deep learning models, DeepPseudo and msPseudo. 3) To enhance interpretability, we propose a pseudo-value-based neural additive model, PseudoNAM, which achieves performance comparable to deep models while offering global and feature-level interpretations. Additionally, we propose the Fair DeepPseudo and Fair PseudoNAM models, incorporating new fairness constraints into a novel pseudo-value-based objective function to ensure equitable and trustworthy survival predictions in the presence of demographic and censoring bias. 4) To enable multi-institution collaboration while preserving data privacy, we introduce federated learning frameworks, FedPseudo for survival analysis and Fedora for competing risk analysis. Furthermore, we introduce a random forest-based federated survival analysis (FSA) framework, FedPRF, to address the communication burden in model exchange and a pioneering fair FSA framework, FairFSA, which integrates fairness through distributionally robust optimization to ensure equitable global survival predictions across clients. We evaluated our approaches on both centralized and decentralized survival datasets, achieving significant improvements over existing methods. These advancements are expected to facilitate better decision-making, optimize healthcare resource allocation, reduce costs, improve treatment interventions and therapy strategies, enhance patient care, and ultimately contribute to improved survival outcomes.
A Robust Federated Learning against Cyber Intrusions to Ensure Data Confidentiality and Model Integrity
(2024/01/01) Ovi, Pretom Roy; Gangopadhyay, Aryya AG; Information Systems; Information Systems
Federated Learning (FL), a type of distributed machine learning, enables collaborative model building among a large number of participants without revealing sensitive data to the central server. In FL, data remains locally on workers’ devices, and every participating worker shares only gradient updates or weight updates with the server after each round of local training. Because of its distributed architecture, FL has limited control over the local data and corresponding training processes. Therefore, it is susceptible to attacks like targeted data poisoning, in which attacked workers use poisonous training samples to train the local model, and the contributions (weight or gradient updates) from these compromised workers poison the global model, resulting in incorrect classifications. Therefore, safeguarding local workers from data poisoning attacks and also detecting those that have already been poisoned are crucial to build a robust federated learning framework. To address it, we propose a prevention strategy, namely `Confident Federated Learning', to prevent the workers from such poisoning attacks. According to our proposed approach, it involves a stratified verification step that validates the label quality of local training data by characterizing and identifying label errors in the local dataset. And experimental results on the MNIST, Fashion-MNIST, and CIFAR-10 datasets suggest that the proposed method can successfully detect the potential mislabeled training samples with above $85$\% accuracy and finally exclude them from local training to prevent data poisoning attacks. However, this strategy is effective with a certain percentage of poisonous local data. And so, in addition to the prevention strategy, we also propose a novel detection method that aims to create a class-wise cluster representation for every participating worker by utilizing the neurons’ activation maps of the local model and finally analyze the resulting clusters to filter out the attacked workers before model aggregation. We experimentally demonstrate the efficacy of our proposed detection strategy in detecting workers affected by data poisoning attacks, along with the attack types, e.g., label-flipping or dirty labeling. Secondly, due to the presence of a gradient sharing protocol as part of the training process, FL is at higher risk in the event of gradient inversion attacks that leak data confidentiality. The most alarming characteristic of this attack is its ability to operate covertly, seamlessly avoiding any degradation to the training performance. Such attacks allow attackers to backtrack from the gradients, eventually reconstructing the private training data. As a countermeasure, we propose a mixed quantization enabled FL scheme that is built upon the concepts of scaler quantization and dequantization to defend against the attack. We experimentally demonstrate the applicability and generalizability of our proposed defense algorithm across computer vision, natural language processing, and audio domains. We utilize eight different datasets of image, audio, and text modalities and cover both iteration and recursion driven gradient inversion attacks. The evaluation of our proposed FL on these benchmark datasets demonstrates the superiority of our method over the existing baseline defense approaches. In addition, our approach can also be considered a communication-efficient federated learning framework because it transforms the high-precision gradients into low-bit precision, resulting in faster training, less transmission bandwidth, and lower communication costs. This dissertation contributes to developing effective defense strategies to defend against these mentioned attack methods in federated learning for maintaining the purity and integrity of the machine learning models, with the aim of safeguarding data, models, and systems.
Regulatory Compliance within the Software Industry: Interpretation of Regulatory Ambiguity as a Compliance Concern
(2024/01/01) Kempe, Evelyn Marie; Sampath, Sreedevi; Massey, Aaron K; Information Systems; Information Systems
Software companies must demonstrate and communicate due diligence toward compliance with applicable regulations and laws within their organizational processes and procedures. The ambiguous phrasing within regulations (i.e., regulatory ambiguities), though, can be challenging for a software developer trying to develop and interpret regulatory compliance requirements for software. Because of this challenge, software organizations or development teams need help communicating and documenting their compliance process during software development. Legal consultants, regulating officials, or compliance auditors can have similar challenges trying to interpret the development work of a software organization and determine if they have applied the requisite amount of due diligence toward regulatory compliance. My dissertation studies a process to assist software developers in interpreting regulatory ambiguities and accomplishes three goals. The first goal is to understand the software industries’ perceptions of compliance through an interview study and survey. The second goal is to observe the reasoning and communication behind interpreting and modeling regulatory ambiguities within a group of software practitioners via a multi-case study. Finally, goal three validates the ambiguity modeling process as useful from an auditor’s perspective through a focus group. The ambiguity modeling process within this work elicits and documents regulatory analysis to support technical compliance decisions and due diligence within a software development process that is reviewable by regulators or external third parties interested in a software organization’s compliance procedures. This approach has advantages for various stakeholders. This process allows software developers to communicate specific instances of ‘gray areas’ in regulation, such as conflicting requirements or unclear terminology, during compliance requirements development so they may receive further guidance and resolution. For auditors assessing organizations for regulatory compliance, ambiguity models demonstrate that software organizations are aware of and discuss their compliance requirements. The models can facilitate meaningful conversations for other stakeholder groups, bridging communication gaps with a software engineering team that can hinder regulatory compliance analysis and development. For the software engineering community, prior research lists challenges with regulatory compliance. My research addresses some of those challenges while promoting compliance communication amongst software stakeholders and assisting in developing a compliance culture within software organizations.
Advanced Semi-supervised Tensor Decomposition Methods for Malware Characterization
(2024/01/01) Eren, Maksim Ekin; Nicholas, Charles; Computer Science and Electrical Engineering; Computer Science
Malware continues to be one of the most dangerous and costly cyber threats to national security. As of last year, over 1.3 billion malware specimens have been documented, prompting the use of data-driven machine learning (ML) techniques for their analysis. However, existing ML approaches face significant barriers that limit their widespread implementation. These challenges include the detection of novel malware, maintaining performance with low quantities of labeled data during training, and classifying malware under class imbalance: a scenario where malware families are unevenly represented in the dataset. This dissertation addresses these shortcomings by introducing three novel semi-supervised ML methods based on tensor decomposition. Our methods are based on dimensionality reduction, hierarchical tensor decomposition, automatic model determination, and feature extraction methods with selective classification or reject-option capability. This "reject-option" capability is a form of self-awareness that allows our models to abstain from making a decision under uncertainty, which in return allows for detection of novel threats. In this dissertation, we describe the foundational concepts underlying our methods and describe the approaches we developed: the Random Forest of Tensors (RFoT), HNMFk Classifier, and MalwareDNA. Additionally, we detail the capabilities of our methods to utilize High Performance Computing (HPC), multi-processing, and Graphical Processing Units (GPUs) for accelerated computation. We showcase our experiments with all three methods where we demonstrate stable task performance under extreme class imbalance, low-quantity of labeled data, and extreme quantities of malware families. We also showcase results when simultaneously classifying benign-ware and malware, classifying malware families, and detecting novel malware families. Our results are compared against state-of-the-art semi-supervised and supervised ML baselines on two datasets. We showcase how our method surpasses the performance of our baselines with a trade-off in increased abstention or reject-option rate.
Toward a Unified, Hierarchical Ontology of Consensus Proofs in Autonomous Distributed Computing
(2024/01/01) Bonyadi, Cyrus Jian; Sherman, Alan T Duan, Sisi; Computer Science and Electrical Engineering; Computer Science
Distributed systems studies the ways in which individual systems behave as a group and is often a result of convergent evolution in many fields of study which seek to understand consensus: how the behavior of systems within a group emerges into singular actions by the group.Practitioners in distributed computing (a subset of distributed systems) often try to integrate consensus mechanisms from other fields, seeking improvements in performance and security properties, but the wide variety of origins also makes these mechanisms difficult to compare. There are no existing works which attempt to reconcile the differences between such consensus mechanisms in a clear, communicable fashion without the requirement for additional tooling (e.g., formal methods)---a limitation which this work seeks to correct. Consensus theory across distributed systems relies on the idea that a collection of autonomous systems may, together, take possible actions with or without the involvement of all members.This idea captures the three primary properties to modern consensus in distributed computing: termination, validity, and agreement. The types of consensus algorithms in distributed computing are characterized by synchrony, permissions or privacy, and fault tolerance. The first portion of this thesis discusses the correlations of synchrony and termination, permissions and validity, and fault tolerance and agreement. The original intention for this work was to remedy the divisions in consensus by developing three hybrid consensus mechanisms which enabled multiple algorithms across the primary properties, respectively.While these hybrid consensus mechanisms would have helped establish a common language for the field, these mechanisms, ironically, could not be proven in the general case without a common language. As a result, the need for a common language could only be fulfilled by a basis for the foundational structure, which this work adapts from the philosophical field of metametaphysics. Therefore, the second portion of this work directly addresses the problem defined in its title by presenting a usable methodology. Finally, the latter portions provide the original intended works, then explain and describe usability of the adapted philosophy, along with potential next steps for formalized standardization.
Design, Synthesis, and Biological Evaluation of Modified Acyclic Fleximer Nucleoside Analogues
(2024/01/01) Thames, Joy Elizabeth; Seley-Radtke, Katherine L; Chemistry & Biochemistry; Chemistry
As the SARS-CoV-2 pandemic left lingering effects worldwide, it remains clear there is a need for the development of broad-spectrum antiviral therapeutics to not only help fight the pandemic at hand but help fight future pandemics to come. With this in mind nucleos(t)ides have had a rich history as antivirals. A commonly employed modification is the use of an acyclic sugar (lack of 2’ and/or 3’ carbons on the sugar moiety), such as that found in Acyclovir (ACV), an FDA-approved drug for herpes simplex virus or the FDA-approved drug Cidofovir (CDV) for CMV retinitis. Research in the Seley-Radtke group focuses on investigating nucleos(t)ide analogues known as “fleximers”, which feature a purine nucleobase “split” into its imidazole and pyrimidine moieties via a carbon-carbon single bond, thus introducing flexibility to the nucleobase. This novel design has endowed the fleximer analogues with potent activity not seen in the rigid parent nucleosides. Combining the acyclic sugar, with the fleximer technology produced a series of first-generation, doubly flexible Flex-ACV analogues. Some of these analogues have exhibited activity against filoviruses, flaviviruses, and coronaviruses. Given the broad-spectrum activity exhibited by these fleximers, optimization of previously synthesized Flex-ACV analogues was carried out for further investigation into the biological effects of these nucleosides. The first-generation Flex-ACV nucleosides were synthesized using modern organic chemistry techniques such as various cross-coupling methods, as well as microwave assisted chemistry. Synthesis of various monophosphate and triphosphate analogues allowed for elucidation of the metabolism and mechanism of action of the first-generation analogues. A new series of Flex-ACV analogues that feature a bond between carbon 5 of imidazole and carbon 5 of the pyrimidine moiety have been designed and synthesized. The newly designed analogues are of interest to compare them to the typical proximal fleximer design and evaluate their antiviral activity as well as the potential to reach new residues in a viral enzyme binding pocket. Overall, the flexible acyclic nucleosides continue to show broad-spectrum antiviral activity and have the potential to become an antiviral therapeutic.
Comparative Analysis of Female Song Structure and Function in Two Temperate and Tropical Oriole Species
(2024/01/01) Moyer, Michelle J; Omland, Kevin E; Biological Sciences; Biological Sciences
Our understanding of the structure and function of elaborate signaling traits, such as coloration or vocalizations, has been dramatically impacted by a long-held bias towards studying these traits in male animals. Many female animals also have functional elaborate signals, but in the past, these traits were not likely to be studied and were often assumed to have only evolved as a consequence of shared genetic architecture with males. Work from our lab has demonstrated that the common ancestor of all songbirds likely had both male and female bird song, but female bird song remains significantly understudied. In tropical species, which are more likely to defend territories year-round and have sexual monomorphism, female song is more common than in temperate-breeding, migratory species. As a result, there is growing interest in the associations of tropical vs. temperate natural history characteristics with the incidence and function of female song. In this dissertation, we compared male and female song in two closely related New World Blackbirds (Icteridae): Orchard Orioles (Icterus spurius), a migratory, temperate-breeding species, and in Puerto Rican Orioles (Icterus portoricensis), a non-migratory, tropical species. We found that female song in Orchard Orioles is significantly reduced compared to male song in acoustic structure, rate, and function. We were able to reliably distinguish between female and male song in the field, and we did not observe female song produced in any consistent functional contexts. Puerto Rican Oriole female song was also acoustically reduced compared to male song, but to a much lesser degree than in Orchard Orioles. We did not find any reliable acoustic metric for distinguishing between female and male Puerto Rican Oriole song in the field, though females did not typically participate in the dawn chorus. However, females did seem to sing more than males sporadically throughout the day. These results suggest that female song in Orchard Orioles is likely under relaxed selection and may be in the process of being lost, which may parallel its loss of elaborate coloration. Female song in Puerto Rican Orioles seems to experience similar selective pressures to male song and is more likely to be functional, which would be predicted in a tropical species with year-round territory defense. To our knowledge, this is the first detailed comparative study of female song in closely related temperate and tropical species.
Surface Modifications of InSe and MXene Two-Dimensional (2D) Materials and their Impact on the Chemical Stability and Interactions with Bacterial Models Relevant to the Environment
(2024/01/01) Sengupta, Shreyasi; Rosenzweig, Dr. Zeev; Chemistry & Biochemistry; Chemistry
Nanosheets are layered two-dimensional nanostructures with thickness in the nano-range, but with properties completely different from the bulk precursors or 3D nanomaterials. Layered materials are defined as solids with strong in-plane chemical bonds, but weak out-of-plane van der Waals interactions. Their unique structural properties enable new functional properties that are important for a multitude of applications. The goal of my research project is to investigate the impact of surface chemistry modifications of 2D materials like InSe and MXenes on their stability and interactions with environmentally relevant bacterial organisms. From a basic science perspective, this study increases the molecular level understanding of the interactions between hydrophobic 2D materials like InSe and amphiphilic adsorbates. The primary choice of our 2D material is InSe, which is an emerging semiconductor 2D material having a direct band gap, already used in broadly distributed nanotechnologies. Usage of InSe and similar 2D materials is expected to grow exponentially in the next decade. However, their unique morphology and ability to produce reactive oxygen species (ROS) could adversely affect human health and the environment. There is an urgent need to address these environmental concerns before the expected exponential growth of their use. Aim 1 of the project is to perform surface modifications on originally hydrophobic InSe nanosheets by physisorption of an amphiphilic molecule and a natural organic matter (NOM) simulant to make them water soluble and understand the impact of this modification on the chemical stability of the nanosheets. Physisorption of NOM components on InSe nanosheets is likely to naturally happen in the environment (natural water) when InSe nanosheets are disposed at the end of devices lifetime, and thus they act as model environmental aqueous solutions. Aim 2 is to investigate the impact of the surface-modified InSe 2D materials soluble in aqueous media on Shewanella oneidensis MR-1, which is an environmentally relevant Gram-negative bacterial cell model. Aim 3 is to investigate the environmental impact of a different highly conducting and water-soluble emerging set of 2D material, MXenes, in unmodified and surface-modified forms with different functional groups as well as fabricated on thermoelectric composite materials, on Shewanella oneidensis MR-1.

Browse

Recent Submissions