150813 research outputs found
Sort by
Augmented intelligence should be good for medicine, if medicine is to remain good for us
No full textThroughout history, the medical community has failed to address health disparities. Augmented intelligence (AI) is poised to cement these structural inequities permanently. The need to establish a triage process that ensures fair and equitable access to medical care, and to consider all patient populations equally researchable, should not overshadow the need to learn how best to exploit AI for furthering medical fairness and equity despite resource limitations. Open discussion of the shortcomings of medical AI, approaching medical AI development, testing, and implementation from a critical ethical perspective, constant testing and analysis of AI outputs, and human oversight in the loop constitute only the first part of ensuring augmented intelligence tools are equitably robust and free of bias
Embedded Computing for Wavefront Control on Future Space Telescopes
No full textFuture space telescopes will use adaptive optics to suppress starlight to directly image and characterize exoplanets. A measurement using this technique may be the first to detect extraterrestrial life in the universe. However, the real-time execution of adaptive optics control algorithms places unprecedented demands on spaceborne processors. Previous work has determined that processing limitations can degrade the achievable contrast and scientific yield of future exoplanet imaging missions. In this work, we quantify the relationship between adaptive optics processing needs and high contrast performance for the Habitable Worlds Observatory (HWO), a mission expected to launch in the 2040s and achieve the 10^-10 contrast necessary to image Earth-like planets around Sun-like stars.
We survey the current landscape of high-order wavefront sensing and control (HOWFSC) algorithms for a future mission like HWO. We parameterize the compute requirements of multiple algorithms through analyses of computational patterns, benchmarks, and problem scaling. In parallel, we assess the capabilities of current and emerging spaceborne processors. We integrate these findings to model processor requirements across several dimensions of telescope design, and we predict whether various processor choices can meet the computational demands of specific HWO configurations. To validate our models, we implement HOWFSC algorithms on representative embedded processors and compare measured performance to predictions. These implementations also reduce risk for spaceflight by increasing the technology readiness level (TRL) of the algorithm–processor pairing to TRL 4.
Given the significant uncertainty in HWO’s eventual design, we extend our deterministic models using Monte Carlo methods to evaluate system performance under uncertainty. We identify key sources of uncertainty and estimate the achievable contrast across a range of system configurations. Our results show that offloading computation to the ground is an important architectural option for most HWO designs. Even under optimistic assumptions, current space processors are insufficient to support the full range of HWO configurations. However, newly developed efficient algorithms substantially reduce the computational burden. Overall, we estimate that current technology has only a 40% probability of supporting HWO’s mission goals without additional architectural innovations. We conclude by recommending combinations of onboard computing, ground offloading, and optical design constraints to help close this technology gap as the mission design matures. In particular, we find that telescope stability and ground-in-the-loop performance are primary drivers of contrast performance, while algorithmic advances such as AD-EFC and onboard compute approaching ground-based GPU performance also provide significant benefits.Ph.D
Modeling Recursion with Iteration: Enabling LLVM Loop Optimizations for Recursive Data Structure Traversal
No full textRecursive algorithms are a natural and expressive way to traverse complex data structures, but they often miss opportunities for optimization in modern compiler infrastructures like LLVM. This thesis explores a novel technique that temporarily transforms recursive traversals into synthetic loop-like structures, enabling existing loop-specific optimizations to apply, before transforming them back. By extending Clang’s semantic analysis and implementing a custom LLVM transformation pass, recursive traversals are initially structured into synthetic loops that can benefit from existing loop analyses and optimizations. After these optimizations are applied, the transformation restores the original recursive semantics, preserving program behavior while incorporating performance gains. Evaluation across custom microbenchmarks shows that while general recursive traversals suffer a modest overhead, workloads designed to benefit specific loop-focused optimizations achieve up to a 30% performance improvement. This demonstrates that even though the approach requires temporarily "misrepresenting" code to the compiler, selective exposure of recursive patterns to loop-based optimization infrastructure is practical and effective. This work establishes a proof-of-concept for compiler transformations that bridge recursion and iteration, paving the way for future systems that better optimize real-world recursive code without sacrificing clarity or maintainability.M.Eng
Assessing concordance between RNA-Seq and NanoString technologies in Ebola-infected nonhuman primates using machine learning
No full textThis study evaluates the concordance between RNA sequencing (RNA-Seq) and NanoString technologies for gene expression analysis in non-human primates (NHPs) infected with Ebola virus (EBOV). A detailed comparison of both platforms revealed a strong correlation, with Spearman coefficients for 56 out of 62 samples ranging from 0.78 to 0.88. The mean and median coefficients were 0.83 and 0.85, respectively. Bland-Altman analysis confirmed high consistency across most measurements, with values falling within the 95% limits of agreement. Using a machine learning approach with the Supervised Magnitude-Altitude Scoring (SMAS) method trained on NanoString data, OAS1 was identified as a key gene signature for distinguishing RT-qPCR positive from negative samples. Remarkably, when used as the sole predictor in a logistic regression model, OAS1 maintained its predictive power on RNA-Seq data from the same cohort of EBOV-infected NHPs, achieving 100% accuracy in distinguishing infected from non-infected samples. OAS1 was also tested in a completely independent held-out test set, consisting of human monocyte-derived dendritic cells (DC) isolated and infected with different strains of the Ebola virus: wild-type (wt), VP35m, VP24m, along with a double mutant VP35m & VP24m, and again demonstrated a 100% accuracy rate in differentiating EBOV-infected from mock-infected samples, confirming its effectiveness as a predictive marker across diverse experimental setups and virus strains. Further differential expression analysis across both platforms identified 12 common genes (including ISG15, OAS1, IFI44, IFI27, IFIT2, IFIT3, IFI44L, MX1, MX2, OAS2, RSAD2, and OASL) that showed the highest levels of statistical significance and biological relevance. Gene Ontology (GO) analysis confirmed the involvement of these genes in key immune and viral infection pathways, highlighting their importance in EBOV infection. RNA-Seq uniquely identified genes such as CASP5, USP18, and DDX60, which are important in immune regulation and antiviral defense and were not detected by NanoString, demonstrating the broader detection capabilities of RNA-Seq. This study indicates a very strong agreement between RNA-Seq and NanoString platforms in gene expression analysis, with RNA-Seq displaying broader capabilities in identifying gene signatures
Large-scale comparison of Fe and Ru polyolefin C–H activation catalysts
No full textWe performed a large-scale density functional theory comparison of polyolefin C–H hydroxylation trends across over 200 Fe and Ru catalysts that are identical except for their metal centers for the radical-rebound conversion of propane to propanol. We observed a strong spin-state dependence: higher-spin states had more favorable metal-oxo formation and isopropanol release in Ru catalysts, while hydrogen atom transfer (HAT) was more favorable in Fe catalysts. While the widely studied metal-oxo formation vs. HAT linear free-energy relationship held for Ru, it was more easily disrupted for Fe. Ru catalysts have a spin-forbidden C–H hydroxylation pathway, while Fe catalysts favor a spin-allowed, intermediate-spin pathway. Calculation of reaction coordinates on representative catalysts corroborated these spin–reactivity trends and showed comparable energetic spans for Fe and Ru analogues, as well as strong Brønsted–Evans–Polanyi relationships for both the metal-oxo formation and HAT steps, motivating expanded study of Fe catalysts
Yeast Display Reveals Plentiful Mutations That Improve Fusion Peptide Vaccine-Elicited Antibodies Beyond 59% HIV-1 Neutralization Breadth
No full textBackground/Objectives: Vaccine elicitation of antibodies with high HIV-1 neutralization breadth is a long-standing goal. Recently, the induction of such antibodies has been achieved at the fusion peptide site of vulnerability. Questions remain, however, as to how much anti-fusion peptide antibodies can be improved and whether their neutralization breadth and potency are sufficient to prevent HIV-1 infection. Methods: Here, we use yeast display coupled with deep mutational screening and biochemical and structural analyses to study the improvement of the best fusion peptide-directed, vaccine-elicited antibody, DFPH_a.01, with an initial 59% breadth. Results: Yeast display identified both single and double mutations that improved recognition of HIV-1 envelope trimers. We characterized two paratope-distal light chain (LC) mutations, S10R and S59P, which together increased breadth to 63%. Biochemical analysis demonstrated DFPH-a.01_10R59P-LC, and its component mutations, to have increased affinity and stability. Cryo-EM structural analysis revealed elbow-angle influencing by S10R-LC and isosteric positioning by S59P-LC as explanations for enhanced breadth, affinity, and stability. Conclusions: These results, along with another antibody with enhanced performance (DFPH-a.01_1G10A56K-LC with 64% breadth), suggest that mutations improving DFPH_a.01 are plentiful, an important vaccine insight
Lanmodulin‐Decorated Microbes for Efficient Lanthanide Recovery
No full textRare earth elements (REEs) are essential for many clean energy technologies.Yet, they are a limited resource currently obtained through carbon-intensivemining. Here, bio-scaffolded proteins serve as simple, effective materials forthe recovery of REEs. Surface expression of the protein lanmodulin (LanM) onE. coli, followed by freeze-drying of the microbes, yields a displayed proteinmaterial for REE recovery. Four REE cations (Y3+, La 3+, Gd3+, and Tb3+) arecaptured efficiently, with over 80% recovery even in the presence ofcompetitive ions at one-hundred-fold excess. Moreover, these materials arereadily integrated into a filter with high capture capacity (12 mg g−1 dry cellweight) for the selective isolation and recovery of REEs from complexmatrices. Further, the proteins in the filter remain stable over tenbind-and-release cycles and a week of storage. To improve the deployability ofthis filter material, a simple colorimetric assay with the dyealizarin-3-methyliminodiacetic acid is incorporated. The assay can beperformed in under 5 min, enabling rapid monitoring of REE recovery andfilter efficiency. Overall, this low-cost, robust material will enableenvironmentally friendly recycling and recovery of critical elements
DexWrist: A Robotic Wrist for Constrained and Dynamic Manipulation
No full textDexWrist is a compliant robotic wrist designed to advance robotic manipulation in highly-constrained environments, enable dynamic tasks, and speed up data collection. DexWrist is designed to be close to the functional capabilities of the human wrist and achieves mechanical compliance and a greater workspace as compared to existing robotic wrist designs. The DexWrist can supercharge policy learning by (i) enabling faster teleoperation and therefore making data collection more scalable; (ii) completing tasks in fewer steps which reduces trajectory lengths and therefore can ease policy learning; (iii) DexWrist is designed to be torque transparent with easily simulateable kinematics for simulated data collection; and most importantly (iv) expands the workspace of manipulation for approaching highly cluttered scenes and tasks. More details about the wrist can be found at: https://sites.google.com/view/dexwrist/home.S.M
Robust resonant anomaly detection with NPLM
No full textIn this study, we investigate the application of the New Physics Learning Machine (NPLM) algorithm as an alternative to the standard CWoLa method with Boosted Decision Trees (BDTs), particularly for scenarios with rare signal events. NPLM offers an end-to-end approach to anomaly detection and hypothesis testing by utilizing an in-sample evaluation of a binary classifier to estimate a log-density ratio, which can improve detection performance without prior assumptions on the signal model. We examine two approaches: (1) a end-to-end NPLM application in cases with reliable background modelling and (2) an NPLM-based classifier used for signal selection when accurate background modelling is unavailable, with subsequent performance enhancement through a hyper-test on multiple values of the selection threshold. Our findings show that NPLM-based methods outperform BDT-based approaches in detection performance, particularly in low signal injection scenarios, while significantly reducing epistemic variance due to hyperparameter choices. This work highlights the potential of NPLM for robust resonant anomaly detection in particle physics, setting a foundation for future methods that enhance sensitivity and consistency under signal variability
Sound2Haptic: A Toolkit for Portable Multi-Channel Haptic Integration Across Multiple Form Factors and Devices
No full textUIST Adjunct ’25, Busan, Republic of KoreaExisting multi-actuator vibrotactile systems often require external hardware such as sound cards and haptic amplifiers, which limits portability and creates complexity for non-technical users. This presents a significant barrier for researchers and designers in fields like human factors and healthcare. We present Sound2Haptic, an vibrotactile toolkit that integrates a sound card and haptic amplifiers into a single device. The toolkit connects to laptops, phones, and XR headsets, enabling portable eight-channel multi-actuator interaction accessible to non-technical users. The toolkit features a novel mechanical design that reduces cross-actuator interference and enables form factor customization. We demonstrate the toolkit’s functional efficacy through psychophysical evaluation across three form factors, and its ease of use through three case studies: (1) a clinical application for tinnitus research (2) a human factors study on speech prosody conducted with human factors researcher, and (3) an exploration of spatial neglect rehabilitation using XR and haptics