62588 research outputs found
Sort by
Full-Low Evaluation Methods for Derivative-Free Optimization
No full textDerivative-free optimization (DFO) addresses the optimization of functions when derivativesare not available or are expensive to evaluate or are subject to noise. In this thesis, we
first introduce Full-Low Evaluation for unconstrained optimization, a novel class of rigorous
methods for DFO with the aim of delivering efficient and robust numerical performance
for functions of all types, from smooth to non-smooth, and under various noise regimes.
These methods are structured around two iteration types: Full-Eval, which is expensive in
function evaluation but performs well in smooth, non-noisy scenarios, and Low-Eval, which
uses less evaluations and is more robust in noisy or non-smooth situations. Our implemented
approach combines finite-difference BFGS with probabilistic direct search (pDS),
showing efficiency and robustness across diverse problem settings. Secondly, we extend
this framework to handle bounds on variables and linear constraints. We derive convergence
results for an instance of this framework that combines finite-difference quasi-Newton
steps with pDS steps. The former are projected onto the feasible set, while the latter
are defined within tangent cones identified by nearby active constraints. We illustrate the
practical performance of our instance on standard linearly constrained problems, that we
adapt to introduce noisy evaluations as well as non-smoothness. In all cases, our method
performs favorably compared to algorithms that rely solely on Full-Eval or Low-Eval iterations.
Thirdly, we assess the use of neural networks as surrogate models to approximate and
minimize objective functions in optimization problems. We show that the best activation
functions for approximating the objective functions of popular nonlinear optimization test
problems are ReLU and SiLU. We also show that neural networks can deliver competitive
zeroth- and first-order approximations at a high training cost. Lastly, we provide evidence
that the performance of a state-of-the-art DFO algorithm can hardly be improved using
surrogate models, including neural networks.</p
Vessel-Supported Tumor Organoids and CRISPR Assays: Innovations in Cancer and Pathogen Detection
No full textIn recent years, patient-derived organoid models have gained prominence for studying cancer biology and personalizing therapies. Unlike 2D cell lines, these 3D organoids faithfully mimic primary tumor features. To further enhance the accuracy and efficiency of tumor modeling, we report a novel approach to expedite preclinical anticancer drug screening using vessel-supported patient-established tumor organoids. Our method involves a standardized and speeding mechanical micro-sectioning process to produce hundreds to thousands of patient tissue fragments from tumorous surgical discard, enabling the fast formation of the tumor organoids. These established organoids are then integrated into a two-layer microfluidic vasculature platform, simulating the tumor microenvironment with flow perfusion-enhanced dynamic culture. The significance of a perfusable vascular network becomes evident during chemotherapy drug screening, as different types of drugs are perfused into the system for various patients. Under this dynamic condition, the dose-dependent effect on viability is less impactful compared to traditional static conditions. The combination of mechanically-established patient-derived tumor organoids with a cost-effective microfluidic vascular system holds great promise for personalized medicine. By accurately replicating patient-specific drug responses in a dynamic tumor microenvironment, this approach could advance preclinical drug screening, thus leading to improved treatments for patients with lung cancer and beyond.The COVID-19 pandemic has presented a significant challenge to the world\u27s public health and led to over 6.9 million deaths reported to date. A rapid, sensitive, and cost-effective point-of-care virus detection device is essential for the control and surveillance of the contagious severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) pandemic. The study presented here aimed to demonstrate a solid-phase isothermal recombinase polymerase amplification coupled CRISPR-based (spRPA-CRISPR) assay for on-chip multiplexed, sensitive and visual COVID-19 DNA detection. The assay targets the SARS-CoV-2 structure protein encoded genomes and can simultaneously detect two specific genes without cross-interaction. The amplified target sequences were immobilized on the one-pot device surface and detected using the mixed Cas12a-crRNA collateral cleavage of reporter-released fluorescent signal when specific genes were recognized. The endpoint signal can be directly visualized for rapid detection of COVID-19. The system was tested with samples of a broad range of concentrations (20 to 2x104 copies) and showed analytical sensitivity down to 20 copies per microliter. Furthermore, a low-cost blue LED flashlight (~$12) was used to provide a visible SARS-CoV-2 detection signal of the spRPA-CRISPR assay which could be purchased online easily. Thus, our platform provides a sensitive and easy-to-read multiplexed gene detection method that can specifically identify low concentration genes.</p
Utilizing In Situ Reflection High Energy Electron Diffraction to Investigate the Structural Transformations of Oxide Thin Films
No full textA custom atomic layer deposition (ALD) chamber equipped with reflection high energy electron diffraction (RHEED) was designed, manufactured, and calibrated to study the structure of surfaces and thin films. The oxides studied include aluminum oxide, hafnium oxide, and gallium oxide with a focus on gallium oxide due to its many polymorphs that can exist depending on the deposition temperature, precursor dosage or chemistry, and substrate structure or orientation. In these studies, amorphous, polycrystalline, and epitaxial films were studied with the RHEED-equipped ALD system and supported by other measurements. The first stage of this project consisted of the design of an ALD chamber with RHEED. The chamber body was chosen to have an inner diameter of 10 in, which ensured that nearby system components did not exceed 250 °C to prevent degradation of the Kalrez o-rings. Flow simulations were also performed to locate any areas that could harbor vortices and set a maximum flow velocity that would ensure laminar flow through the chamber. Considerations for RHEED incorporation were also addressed such as understanding how pressure will affect the quality of RHEED by evaluating the mean free path as a function of temperature. Other considerations such as maximum viable chamber pressure to maintain less than 1x10-6 Torr in the RHEED electron source were also discussed. Calibration was also required to determine the camera length of this custom system using a known sample, and the wafer surface temperature by observing the temperature curve of a thermocouple bonded to a wafer surface. This system was used to deposit and observe the structural transformations of thin, oxide films described in this work.
Aluminum oxide on c-plane sapphire, hafnium oxide on silicon, and gallium oxide on silicon were initially investigated to test the capabilities of the newly manufactured chamber. A 3.3 nm thick, amorphous alumina film was deposited ex situ then heated to 640 °C with RHEED irradiation. There were no definitive indicators that the film crystallized, but a change in the observed RHEED pattern indicate that a structural rearrangement may have taken place. Three, approximately 30 nm, amorphous hafnia films were deposited on Si chips ex situ and then heated to 675 °C while being probed using RHEED. RHEED indicated that crystallization occurred between 410 and 415 °C which is similar to crystallization temperatures recorded in literature. The observed ring pattern was indexed. Approximately 5 nm of gallium oxide was deposited on Si ex situ and annealed under RHEED irradiation. A bright halo was observed beginning as early as 190 °C and becoming more prominent as the temperature increased to 575 °C. The RHEED ring pattern observed was unfortunately unable to be indexed. These preliminary measurements proved that this custom-built chamber was capable of observing the polycrystalline crystallization 0f ultrathin ALD films.
Thin films of gallium oxide were deposited in an amorphous (low temperature) and, separately, epitaxial form (high temperature) to evaluate how their crystallinity impacted their final structure. The films were then annealed at ultra-high vacuum (UHV) under RHEED observation. The film deposited at 303 °C was crystalline while the one deposited at 217 °C was amorphous and crystallized during the post-deposition anneal. All samples exhibited a streak pattern as observed by RHEED, so the widths between the streaks were converted to d-spacings and the changes in these d-spacings were observed during the deposition process (for films that deposited crystalline) and the annealing procedure. These d-spacings were used to determine the in- and out-of-plane orientation relationship and make observations about the potential strain of each film. Ex situ x-ray diffraction was used to confirm the epitaxial orientation of films and cross-sectional transmission electron microscopy evaluated the film/substrate interface and overall film structure.
A series of amorphous gallium oxide films were deposited on sapphire to investigate how varying film thickness influences crystallization behavior and final film structure. Films between 2 and 9 nm were deposited and annealed to 675 °C for this investigation. The streak pattern widths were again converted into d-spacings to evaluate the in- and out-of-plane epitaxial relationship between the film and the substrate. To evaluate film structural stability, a series of films were annealed one or more times in vacuum. Complete desorption of films less than 3 nm was observed upon a single annealing procedure, so it was investigated that desorption could be suppressed by lowering the maximum temperature, annealing at a higher pressure, or introducing a small partial pressure of oxygen into the chamber during annealing. Structural differences were observed between 6 and 9 nm thick films where thicker films more closely resembled films that deposited epitaxially. Ex situ x-ray photoelectron spectroscopy was used to determine the thickness of all films.
It was observed in the previous experiments that the surface of gallium oxide thin films can change with increasing film thickness, annealing temperature, or annealing ambient gas environment. RHEED pattern simulations were performed in an effort to understand how surface symmetry or the presence of twins can influence the observed RHEED pattern. Overall, this work has shown that RHEED is a useful tool to understand the structural nuances that occur within thin films during epitaxy, crystallization, and annealing under varying atmospheres.</p
How Does Generative AI Usage Affect the Coding Performance of Developers?
No full textThis project examines how Generative AI tools impact developers\u27 coding performance, using empirical analysis and proprietary data from an IT organization.Generative AI (GenAI) tools, such as GitHub Copilot, have emerged as transformative technologies in software development, offering the potential to enhance coding efficiency through real-time suggestions. However, the impact of GenAI on developers\u27 coding performance remains underexplored. This study investigates the effects of GenAI usage on coding productivity by leveraging a difference-in-differences (DID) analysis of 27 weeks of proprietary data from a mid-sized global information technology organization. Results reveal a significant increase in the number of user stories completed per week among developers using GenAI tools. Additionally, we examine the moderating role of developers\u27 working experience, identifying nuanced mechanisms by which GenAI tools influence performance. These findings contribute to theoretical advancements in understanding GenAI\u27s role in software development and offer practical insights for optimizing AI tool adoption in professional settings. The paper, sharing the same title as this project, will be presented at the Hawaii International Conference on System Sciences (HICSS) in January 2025, a premier academic research conference on information systems.</p
Physics-Guided AI Towards Better Diagnosis on Heart Diseases
No full textWe are developing physics-guided machine learning techniques for cardiovascular systems, to obtain a personalized digital twin model for heart diseases.Heart disease, characterized by changes in vascular, valvular, and ventricular systems, has been the leading cause of death in US since 1950. These diseases often happen acutely, making early detection essential. This interdisciplinary project aims to transform the clinical approach to early stage heart diseases, which is still dominant by traditional methods, through an AI-based approach. Our goal is to integrate data from wearable devices, mechanical measurements, and physics-based modeling, towards a personalized digital twin model for heart diseases. To this end, the project has three sub-tasks. First, a physically interpretable constitutive modeling approach will be developed, to capture the change of mechanical responses and the underlying mechanisms for cardiovascular tissue degradation. Second, an attention-mechanism-based approach will be designed, to infer the correlation between multiphysics data and the corresponding tissue model. Third, an efficient surrogate model will be constructed, enabling fast inference to assist in real-time monitoring. As the long-term goal, this personalized digital twin model is anticipated to provide early risk assessments, while also offering insights into cardiovascular mechanics that could lead to new treatment strategies. In the past semester, our team has made progresses on the first and the second sub-tasks. For the first sub-task, a peridynamic neural operator model has been developed, which constructs the constitutive law as well as the underlying fiber orientation from loading-response measurements. To demonstrate the applicability of our approach, we apply the HeteroPNO in learning a tissue model and fiber orientation field from DIC measurements of a heart valve leaflet specimen. For the second sub-task, we have developed an attention-mechanism-base approach to extract global prior information from mechanical measurements of multiple tissue specimens, and provide a generalizable foundation model to new and unseen tissues. As the ongoing work, we are extending this model to multiphysics and multimodal data.</p
Living Simultaneously in Two Different Worlds
No full textBlack students who attend predominantly white institutions before college have a complex educational experience because of the systematically racist practices that are in place within the American school system – students\u27 lives and identities are complicated as they try to get an education. This study focuses on the elite preparatory school environment (places that have been historically for the rich and white), where many Black students integrate into. First, this paper investigates the historical precedents that are the consequence of an unfair and unequal education system. Then, through a series of interviews with Black alumni of prestigious secondary schools this study explores the lives Black students experienced. This research paper examines the contrasting and contradictory environments the students live in, showing how it infers with Black students\u27 educational journey. This research moves past simplistic narratives of racial trauma and disadvantages, instead it analyzes the agency these students have and the ways they navigate these spaces emotionally and strategically
Mother Your Monsters
No full textReproductive labor, the work that people put into raising children and keeping households, is one of the most under-appreciated forms of labor. Monster stories, specifically Frankenstein and Dracula, provide a clear representation of the danger of life without reproductive labor and reveal how incredibly important this work really is. By engaging with other scholars who have examined the text in similar ways and using evidence from the novel, I will set up Frankenstein as the framework for how a gothic novel understands reproductive labor. Then, I will demonstrate how Dracula allows for the same understanding, despite rarely being analyzed in this way. In conclusion, I will examine how Frankenstein and Dracula, despite having been written 200 years ago, provide a depiction of reproductive labor that must be recognized. They demonstrate how vital it is to society, families, and humanity itself. Without reproductive labor, people become monstrous, and society falls apart. These texts depict reproductive labor\u27s importance and reveal how we as a society must start giving the people who undertake that labor their due respect
Flood Risks of Cyber‐Physical Attacks in a Smart Storm Water System
No full textThe rise in smart water technologies has introduced new cybersecurity vulnerabilities for water infrastructures. However, the implications of cyber‐physical attacks on the systems like urban drainage systems remain underexplored. This research delves into this gap, introducing a method to quantify flood risks in the face of cyber‐physical threats. We apply this approach to a smart stormwater system—a real‐time controlled network of pond‐conduit configurations, fitted with water level detectors and gate regulators. Our focus is on a specific cyber‐physical threat: false data injection (FDI). In FDI attacks, adversaries introduce deceptive data that mimics legitimate system noises, evading detection. Our risk assessment incorporates factors like sensor noises and weather prediction uncertainties. Findings reveal that FDIs can amplify flood risks by feeding the control system false data, leading to erroneous outflow directives. Notably, FDI attacks can reshape flood risk dynamics across different storm intensities, accentuating flood risks during less severe but more frequent storms. This study offers valuable insights for strategizing investments in smart stormwater systems, keeping cyber‐physical threats in perspective. Furthermore, our risk quantification method can be extended to other water system networks, such as irrigation channels and multi‐reservoir systems, aiding in cyber‐defense planning. , Key Points We proposed a mathematical framework for evaluating flood risks of cyber‐physical attacks in a smart stormwater system False data injection can maliciously increase inflow and reduce the outflow of a targeted detention pond in a smart stormwater system Additional flood risks caused by false data injection are higher with smaller, more frequent storm
Co‐rumination between friends
No full textDespite its implications for adjustment, little is known about factors that support co‐rumination in friendships. The current multi‐method, longitudinal study addressed this question with 554 adolescents ( M age = 14.50; 52% girls; 62% White; 31% Black; 7% Asian American) from the Midwestern United States in 2007–2010. Adolescents were observed talking about problems with a friend and reported on their outcome expectations for problem disclosures, relationship provisions during problem talk, and problem perceptions after problem talk. Participants reported on outcome expectations again 9 months later. Results indicate that the positive relationship provisions associated with co‐rumination may outweigh negative problem perceptions in predicting adolescents\u27 outcome expectations for problem disclosures over time. Implications for the potentially reinforcing nature of co‐rumination are discussed
Solvent‐cast <span style="font-variant
No full textThe biochemical and physical properties of a scaffold can be tailored to elicit specific cellular responses. However, it is challenging to decouple their individual effects on cell‐material interactions. Here, we solvent‐cast 3D printed different ratios of high and low molecular weight (MW) poly(caprolactone) (PCL) to fabricate scaffolds with significantly different stiffnesses without affecting other properties. Ink viscosity was used to match processing conditions between inks and generate scaffolds with the same surface chemistry, crystallinity, filament diameter, and architecture. Increasing the ratio of low MW PCL resulted in a significant decrease in modulus. Scaffold modulus did not affect human mesenchymal stromal cell (hMSC) differentiation under osteogenic conditions. However, hMSC response was significantly affected by scaffold stiffness in chondrogenic media. Low stiffness promoted more stable chondrogenesis whereas high stiffness drove hMSC progression toward hypertrophy. These data illustrate how this versatile platform can be used to independently modify biochemical and physical cues in a single scaffold to synergistically enhance desired cellular response