33760 research outputs found
Sort by
SOCIOECONOMIC RISK AND RESILIENCE FACTORS FOR CHILDREN’S EARLY COGNITIVE AND SOCIOEMOTIONAL DEVELOPMENT
No full textChildren’s early cognitive and socioemotional (SE) skills constitute their levels of school readiness, a key topic for research and policy due to their relation to children’s long-term academic and career success as well as mental health. This dissertation comprises three empirical studies that investigate socioeconomic risk and resilience processes relating to early cognitive and SE skills among racially, ethnically, and socioeconomically diverse preschoolers and kindergarteners. Paper 1 examines whether parenting relationships acted as a protective moderator reducing the effects of low socioeconomic status (SES) and household chaos on children's executive functioning (EF) and SE skills. This pre-registered study used data from 83 families with a preschooler or kindergartener. SES was associated with children’s EF skills measured by lab-based tasks, but not parent survey-based measures of EF or SE skills. Household chaos was related to children’s SE skills, but not EF. We found evidence of moderation, but not in the expected direction. Thus, this study failed to find evidence of parenting relationships as protective against risk from SES and household chaos for children’s EF and SE skills. However, parenting relationships were more strongly related to children’s survey EF and SE skills than either risk factor, indicating the importance of parenting relationships for children’s development across risk levels. Paper 2 is a pre-registered study using kindergarten data from the nationally representative Early Childhood Longitudinal Studies-Kindergarten: 2011 cohort (ECLS-K:2011) dataset to examine profiles of school readiness resilience amidst socioeconomic risk. We ascertained profiles of resilience in math, reading, and SE skills among kindergarteners exposed to socioeconomic risk, and examined whether contextual factors varied across profiles. Resilience was common among the five profiles identified. Two exhibited mostly resilience (Strong and Strong at School/Average Home SE), with average outcomes at or above the national means, while two profiles exhibited more vulnerability (Strong Math/Vulnerable Reading & SE and Vulnerable), with average outcomes mostly below the national means. Children in the Strong profile had the most socioeconomic resources across five measures, while children in the Vulnerable profile experienced the most socioeconomic risk across all measures. The other three profiles consistently fell in between on all contextual measures. Results add nuanced insights into contextual factors characterizing distinct resilience profiles with unique strengths and weaknesses. Paper 3 investigated the validity of dyadic language measurements across settings. Conversational turns (CTs) among 46 families were measured across two contexts—lab-based interactions and naturalistic recordings. This study examined convergent validity by probing relations between CT measures and contextual factors (SES, race/ethnicity, and stress). To examine concurrent validity, we analyzed relations between each CT approach and standardized assessments of children’s language skills. We found no relations among the three language measures or between either CT approach and social context variables. However, children’s race, ethnicity, and especially socioeconomic status were moderately to very strongly related to standardized language scores. Together, these three studies elucidate socioeconomic risk and resilience factors in early childhood development of cognitive and SE skills. Results have widespread implications for education, social, economic, and health research and policy
Spin Dynamics, Pulsed Decoupling, and Deep Reinforcement Learning for Quantum Sensing
No full textQuantum sensing faces challenges in moving from proof-of-concept demonstrations to practical technologies, with discrepancies between idealized theoretical frameworks and experimental realities presenting a key obstacle. This dissertation explores these nuances — with a focus on nitrogen-vacancy (NV) centers in diamond — through four projects spanning experiment, theory, simulations, and algorithmic optimization.
We characterize Ramsey envelope modulation effects in 15NV diamond magnetometry, showing that magnetic field misalignments produce envelope effects that degrade sensitivity. Next, we study the breakdown of Average Hamiltonian Theory (AHT) in experimental regimes, introduce exact methods to calculate a sensor response to a target signal that are valid beyond AHT, and establish symmetries that guarantee AHT convergence. With Ensemble Cluster Sampling (ECS), we address overfitting in dynamical decoupling by training algorithms on heterogeneous parameter distributions rather than idealized systems. Finally, we present TEMPO, an open-source Python package for accessible pulse sequence simulations.
These studies underline the need for continued collaboration between quantum sensing theory and laboratory applications, while maintaining an eye on modern algorithms and software
Measurement, Simulation, and Compact Modeling of Complex Electron Dynamics
No full textSystems containing large numbers of electrons can exhibit surprisingly complex and rich dynamics. In this dissertation, we ask: What is the minimum necessary detail in measurement or data-driven modeling and simulation to capture complex dynamics manifesting from these systems? To answer this, we integrate experiment, simulation, and theory to understand their complex dynamics. In this dissertation, we examine two such systems: (i) a superparamagnetic tunnel junction (SMTJ) and (ii) charged-particle beam dynamics.
We first consider the deterministic-stochastic behavior of a constant-current driven SMTJ, where we create a measurement-driven overdamped Langevin model capturing statistical properties of the device. We show both how this model captures device statistics across time scales and how it can be refined to capture higher-order behavior.
We next examine the centroid motion of a charged-particle beam and propose a method for understanding and predicting it using an interpretable, data-driven approach whose output is directly identifiable to terms in underlying low-dimensional evolution equations. We derive the evolution equations solely on the basis of data—with no recourse to an underlying first principles model. We compare and contrast our methodology with both a machine learning technique and a first principles model, and we show that we can learn interpretable equations for nonlinear beam dynamics at lower computational cost while achieving comparable accuracy.
Lastly, we investigate the phase space evolution of a charged-particle beam. Accurate knowledge of the phase space at beam creation is crucial for understanding and predicting beam dynamics. We measure a velocity space modulation to initialize the phase space of first principle simulations and capture beam statistics and internal beam structure—resembling a cruciform—with high fidelity. This contrasts with both employed first principles models—which do not account for beam structure as they assume a uniform beam cross section—and simulations using ideal phase space distributions. Finally, we demonstrate sensitivity to beam and lattice parameters varied within experimental measurement error
IC DESIGN FOR AGE-RELATED TIMING DEGRADATION USING MACHINE LEARNING
No full textTiming constraints are a major factor in modern IC design.In each clock cycle, data signals must leave one register, travel through the combinational logic of the design, and arrive at another register within a narrow window to satisfy both setup and hold time constraints.
Therefore, static timing analysis (STA) is a major step in the design process.
Designs often have very little margin in their timing, and many design revisions may be needed to ensure all paths' delays are within acceptable ranges.
However, traditional STA considers the timing only of freshly manufactured circuits.
As circuits age, effects such as negative bias temperature instability (NBTI) and hot carrier injection alter the timing properties of the device.
For devices with a long expected service time, it may not be possible to neglect these effects when considering device timing constraints.
In this dissertation, we investigate the design-level ramifications of NBTI and use machine learning (ML) to predict the end-of-life timing information for the design.
We then explore design techniques that use this model to improve resilience against timing degradation.
We begin our work by evaluating the degree of NBTI degradation over the lifetime of an integrated circuit (IC).From our simulation data, we observe that the increased delay due to NBTI is significant compared to the total path delays.
The traditional approach to protecting the design from errors caused by this additional delay is guardbanding, which anticipates a fixed amount of degradation when setting timing constraints for the designer.
This offsets NBTI by artificially reducing the delay allowed along timing paths, effectively reducing the clock period available to the designer.
However, NBTI does not age the design uniformly, and guardbanding is overly pessimistic since it assumes the worst-case degradation for all paths.
To remove some pessimism and increase the delay budget available to designers, we propose a machine learning technique to predict the effects of NBTI in a design.
While existing approaches to modeling NBTI require circuit simulations for each gate in the design, our technique provides designers with end-of-life path delay data at a speed that can be integrated into design automation tools.
Once path-specific NBTI degradation data is available to design tools, new algorithms can be developed which are less pessimistic than guardbanding, which assumes worst-case degradation throughout the design.
The remainder of this dissertation is devoted to exploring applications of our machine learning technique.
First, we investigate the feasibility of planned obsolescence using NBTI as a failure mechanism.
Then, we develop an NBTI-aware placement tool by integrating our machine learning technique into an open-source design tool.
Planned obsolescence refers to techniques used by companies to reduce the performance or cause the failure of their products after a certain period of time has passed, usually after new products have been released.The device failures encourage previous customers to purchase replacements, increasing the demand and sales of new products.
Often, planned obsolescence is implemented in software updates that intentionally slow performance or decrease battery life.
This gives the company direct control over when the degradation begins and even which devices are affected.
However, this approach is potentially detectable to consumers, who may notice a sudden decrease in performance and investigate the cause.
We describe another approach to planned obsolescence which would allow companies with an obsolescence schedule at design time to introduce performance penalties in a more natural way.
Inserting delay buffers into the netlist at design time counteracts the guardband in the circuit timing, resulting in timing violations caused by natural age-related degradation.
This approach to planned obsolescence is more difficult to detect, since it is embedded into the original netlist and follows the same trends as natural aging.
Many design algorithms for various stages of the IC design process utilize timing data from STA.However, STA timing data represents circuits when they are freshly manufactured, which is when they are furthest from failure.
We explore the use of end-of-life timing in design algorithms, specifically for use in timing-driven global placement.
Global placement is a vital step in the design process which allocates standard cells and large macros across the chip canvas.
The placement solutions must place cells with connecting nets near each other to allow the device to meet timing, but there is a limit to how dense any region of the canvas can be.
We utilize our machine learning technique to provide end-of-life timing data for the longest critical paths to a standard placement tool.
Without modifying the underlying placement algorithm, we are able to produce NBTI-aware designs adapted to the anticipated NBTI degradation
Robustness, Detectability, and Data Privacy In AI
No full textRecent rapid advancements in Artificial Intelligence (AI) have made it widely popular in various applications, ranging from autonomous systems to multimodal content generation. However, these models are vulnerable to several security and safety problems. These pitfalls in AI models can potentially allow attackers to jailbreak systems, enabling harmful tasks and leaking sensitive information. As AI models are increasingly deployed in sensitive applications, such as autonomous robots and healthcare devices, the importance of AI safety is growing. To address these issues in today’s AI systems, it is critical to study their vulnerabilities. In this thesis, we explore fundamental aspects of AI safety and security, including robustness, detectability, and data privacy. We analyze challenges posed by adversaries to contribute toward enhancing AI safety and security. First, we examine the robustness of discriminative models, identifying blind spots in computer vision models. We discover the existence of high-confidence star-like paths in the input space, where models consistently predict the same output label despite the addition of significant noise. We also develop provable robustness certificates for streaming classification models. Next, we analyze the challenges in detecting AI-generated content across text and image modalities. Through empirical and theoretical studies, we find that detecting AI content is a difficult task, especially as generative models improve over time. We demonstrate attacks on a wide range of text and image detectors, showing that adversaries can efficiently break these systems to induce both type-I and type-II errors. We then investigate the robustness of generative Large Language Models (LLMs), focusing on adversarial attacks and hallucinations. Our fast, beam search-based adversarial attacks can compromise LLMs in under a GPU minute. We also demonstrate that multi-modal LLMs have increased attack surface and are more vulnerable to adversarial attacks in the real world. These attacks can jailbreak, induce hallucinations, and carry out privacy attacks on LLMs. To mitigate these risks, we propose a suite of methods for detecting hallucinations in both white-box and black-box settings by analyzing the internal dynamics and output probabilities of LLMs. Lastly, we address user privacy concerns by proposing a method to create unlearnable datasets through data poisoning. This method, which is model-agnostic and computationally efficient, ensures that datasets used without user consent become unlearnable by AI systems. In summary, this thesis analyzes various attacking frameworks to understand the robustness and security vulnerabilities of AI models. We hope the insights provided by this work will be useful in designing safer, more secure AI systems
The Effects of Climate on Global Changes in Cropland Extent: Historical Patterns and their Implications for the Future
No full textClimate change is altering the global agricultural landscape, impacting land suitability and driving changes in yield, management, profitability, and even land use. Climate-driven changes in cropland area have the potential to greatly impact food security, environmental health, carbon storage, and climate; however, the magnitude and scale of cropland responses remains poorly characterized. This study combines remote sensing cropland data products based on MODIS and Landsat with ERA5 climate reanalysis data to evaluate how cropland extent changed in regions of changing climate from 2001-2018, and then applies the derived climate-cropland relationships to estimate future climate pressure on cropland from 2015-2050. Over the study period, global cropland area negatively correlated to temperature, with cropland area decreasing by 3%-4% per °C warming. Precipitation, vapor pressure deficit, and potential evapotranspiration all also significantly correlated to cropland change. These relationships were dependent both on the underlying climate and on the rate of change of the cropland; croplands in transition were more strongly correlated to climate factors. After the key linear cropland-climate relationships were identified, they were further stratified by crop type, management (irrigation and fertilization rates), and economic classification. The coefficients of these linear relationships were then used to calculate the climate pressure, or the estimated potential land-use change if the observed partial temperature and precipitation effects were to hold true from 2015-2050 under the high emissions Shared Socioeconomic Pathway (SSP) scenario 5-8.5. Precipitation exerted strong pressure at sub-regional scales, but had little cumulative global impact, since projected trends were highly localized and relatively balanced between gain and loss. Temperature, on the other hand, exhibited a strong negative pressure over global cropland, amounting to a global pressure of reduction -2.5% of baseline 2015 cropland between 2015-2050. This was comparable in magnitude but opposite in direction to the Integrated Assessment Model-projected cropland change included in the Land Use Harmonization 2 (LUH2) SSP5-8.5 scenario, which projected a global increase of 3.7% baseline area per decade, without considering the impacts of climate on crop production. Though the partial temperature effects identified do not account for compound effects from other climate changes or socioeconomic factors, they nevertheless indicate a strong temperature pressure on cropland change that may not be captured by current integrated modeling systems. This study identified regions where climate pressures may significantly alter cropland area change projections, and highlights the importance of considering climate in the evaluation of agricultural land-use change. Better accounting for the observed role of climate changes, and particularly temperature, on cropland area change will be critical to improve future cropland projections
Computational identification of PIWI associated RNA sources using novel small RNA and evolutionary analyses
No full textPIWI-interacting RNAs (piRNAs) form an essential component in the defense mechanismsagainst transposable elements (TEs), ensuring genomic stability in germ cells across animal
species. PiRNAs are small non-coding RNAs that function by guiding PIWI proteins towards
target transcripts using sequence complementarity. PiRNA pathways are essential for fertility.
Understanding these diverse small RNAs and identifying their longer precursors requires the
development of specific computational tools. My thesis work introduces ”piRNA Cluster Builder”
(PICB), an R package developed to identify piRNA generating genomic intervals. PICB address
the challenges of short read mapping and integrate multimapping reads effectively. PICB employs
a nuanced approach to handle unique and multi-mapping alignments differently, ensuring the
detection of piRNA clusters and the accurate identification of cluster boundaries. This method
consists of three-layer analysis unique mapping alignments to generate seeds, primary multimapping
alignments to define cores, and secondary multimapping alignments to finalize clusters.
Using PICB, this study examines the piRNA populations in various model organisms including
Drosophila melanogaster, Schmidtea mediterranea, and mammals Mus musculus and Mesocricetus
auratus. In Drosophila, PICB reveals distinct characteristics of piRNA clusters in ovarian somatic
sheath cells compared to germ cells, highlighting the complexity and specialization of piRNA
pathways. Analysis of S.mediterranea demonstrates the presence of an active ”ping-pong” mechanism
and a high level of strand purification in piRNA clusters, which contrasts sharply with the relatively
homogeneous strand productivity observed in Drosophila ovaries. In mammals, the piRNA
populations in mouse and hamster testes during different developmental stages show different
regulatory mechanisms as revealed by PICB. Notably, pachytene piRNA clusters in mice exhibit
a varying degree of transposon content and distinct patterns of piRNA biogenesis compared to
prepachytene clusters. The comparative analysis provides insights into the evolutionary dynamics
of piRNA clusters, shedding light on conserved regulatory themes and lineage-specific adaptations.
Predicted evolutionary feature has been independently verified using RNA sequencing data confirming
the analysis.
Through supporting cross-species comparisons, PICB paves the way for a deeper understanding
of the piRNA pathway evolutionary and functional diversity. The insights gained from these
comparative studies hold potential implications for improving RNA therapy and understanding
the broader impact of non-coding RNAs in gene expression regulation.
This dissertation includes the following supplementary materials: spreadsheet of piRNA
cluster genomic coordinates; two large scale phylogenetic trees
Vaginal bacteria-derived extracellular vesicles diffuse through human cervicovaginal mucus to enable microbe-host signaling
No full textThe composition of the vaginal microenvironment has significant implications for gynecologic and obstetric outcomes. Where a Lactobacillus-dominated microenvironment is considered optimal, a polymicrobial environment is associated with increased risk for female reproductive diseases. Recent work examined bacteria-derived extracellular vesicles (bEVs) as an important mode of microbe-host communication that may influence reproductive outcomes. However, in order to communicate with female reproductive tissues, bEVs must penetrate the protective cervicovaginal mucus barrier. We demonstrate increased diffusion of bEVs compared to whole bacteria. Additionally, we evaluate the uptake of bEVs by, and the resulting effects on, human vaginal epithelial, endometrial, and placental cells, highlighting potential mechanisms of action by which vaginal dysbiosis contributes to gynecologic and obstetric diseases. Taken together, our work demonstrates the ability of bEVs to mediate female reproductive outcomes and highlights their potential as therapeutic modalities for treating dysbiosis and dysbiosis-associated diseases in the female reproductive tract
PALIMPSESTIC BLACKNESS: MEMES, AESTHETICS AND RELATION
No full textThis dissertation focuses on the relational aesthetics of the meme to consider more broadly the history of blackness and relation. A meme can represent a sentiment or an idea, often through mechanizing humor, play, and irony. In considering the meme, I am interested in how black visual and sonic performances attend to the “feltness” or relatability of blackness. I argue that the meme, its forms, and its movements can be understood within a larger genealogy of black cultural forms. Like many genres in black musical productions, ranging from jazz, blues, Motown, gospel, R&B, and hip-hop, the meme is a practice of call-and-response. Virality is dependent upon the popularity of the image being used. The idea and emotionality of the image is predicated on the image being used. I look to contemporary black art and its expressions of abstractions and surreal articulations to situate the meme as an aesthetic form central to how we communicate and relate. To this end, I examine the performance art of Adrian Piper and the video works of Arthur Jafa. Arthur Jafa’s work in Love is the Message, The Message is Death(2016) and The White Album(2018) calls attention to affective attachments around the project of racialization and considers the logic of enjoyment and empathy through how blackness exists as a public feeling. Piper’s works–Calling Cards 1 & 2(1986-1990), Funk Lessons(1982-1984) and Thwarted Projects, Dashed Hopes, A Moment of Embarrassment(2012)– underscore the various assumptions of the project of racialization and gendering and how ultimately, these assumptions inhibit relationality. Furthermore, her works attend to the affective dimension of race through mechanizing irony, play, discomfort, shock, shame, and absurdity in embodied and relational performance. In Piper and Jafa’s works, I am especially interested in how they privilege the experiential instead of just the ontological. I contend that by studying the aforementioned artist’s works through memetic aesthetics, we can analyze the performativity and contingency of identity
THE DEVELOPMENT OF HEALTH LITERACY AMONG SECOND GENERATION NIGERIAN AMERICANS
No full textBackground: Racial or ethnic minority immigrants living in the U.S. have an 81% higher prevalence of experiencing limited health literacy than individuals belonging to racial ethnic minorities born in the U.S. and other US-born citizens. Although the children of immigrants are impacted by the health behaviors of their parents, health literacy–related research about Nigerians living in the U.S. has focused on first-generation immigrants. In addition, most of the studies have focused on how health information is used to take actions, with almost no study on how individuals learn to find and understand health information. This study used the Cultural Ecology Model (CEM) constructs to assess how second-generation Nigerian Americans learn to find, understand, and use health information (become personally health literate).
Methods: Qualitative in-depth interviews were conducted on Zoom with 20 second-generation Nigerian Americans between the ages of 18 and 24 recruited from several churches in the Metro-Atlanta area of Georgia and by referrals. The thematic data analysis deployed the CEM constructs of environment, culture and technology to examine themes and gather insights about how second-generation Nigerian Americans become personally health literate.
Results: Findings indicate that second-generation Nigerian Americans describe their health literacy experience as “figuring it out,” and that individual characteristics outside of the themes identified in the CEM contribute to health literacy. In the process of "figuring it out", technology played a crucial role for respondents in the study. It serves as a primary source of health information for many individuals, particularly through social media platforms, with TikTok being particularly popular. The extensive network of medical professionals and a strong focus on medical education in the Nigerian community, significantly contribute to improved health literacy. The study also highlights a shift in communication dynamics between second-generation Nigerian Americans and first-generation Nigerian immigrants. Peer-to-peer exchanges have become essential for navigating health literacy, especially concerning mental health issues.
Conclusion: Results solidify the role of Culture, Environment and Technology in the development of health literacy. Specifically, findings highlight the strong role that technological resources and peer-to-peer interactions play in “figuring it out” – a description of the health literacy journey of second-generation Nigerian Americans