搜索结果：Kentucky

This paper investigates knowledge distillation from a large reasoning model (DeepSeek-R1) to a compact student model (Qwen2.5-7B). Using historical problems from the John O'Bryan Mathematics Competition at Northern Kentucky University (2011-2025), we build a Chain-of-Thought (CoT) training corpus through a dual-agent framework. The dataset is used to fine-tune the student model with Low-Rank Adaptation (LoRA) on Apple Silicon hardware using the MLX framework. The base Qwen2.5-7B model achieves 64.67% accuracy on competition problems, while the DeepSeek-R1 teacher achieves 91.40%. An initial 1,000-iteration training run revealed severe overfitting, with validation loss reaching a minimum at iteration 200 before rising steadily. Based on this finding, we ran five independent training runs each limited to 200 iterations with varied random seeds to assess result stability. Across these five runs, the fine-tuned student model achieves a mean accuracy of 69.43% (std dev 0.17%) on the competition dataset, a 4.76 percentage-point improvement over the base model, and generalizes to 73.1% (std dev 0.18%) on the MATH-500 benchmark. We further study how response length affects answer quality a

We introduce an in-domain supervised pipeline designed to counter the out-of-distribution performance drop that hampers supervised biomedical NLP models, a problem observed when models trained on pathology reports are moved across cancer registries. Our contribution is a reproducible recipe for training a supervised classifier from routinely collected cancer registry data. It describes how to build the in-domain training set and a production-matched holdout, and to choose operating points that keep the false-negative rate (FNR) very low while keeping reviewer workload manageable. The pipeline standardizes data curation with facility-stratified sampling and separate handling of reports linked to registry cases, and includes a blinded manual audit to estimate positive-case prevalence and label noise. On a 418k-report holdout set, the Kentucky model achieved FNR 0.003 and false-positive rate (FPR) 0.097, improving over the Seattle-trained MOSSAIC OncoID baseline (FNR 0.010, FPR 0.183) and raising F1 from 0.860 to 0.922. In a blinded manual review of 600 reports, estimated positive prevalence declined from 0.500 to 0.398, indicating substantial label noise with errors concentrated in r

Mountainous terrain is increasingly being measured and mapped by airplane-based LiDAR (Light Detection and Ranging) techniques, but the accuracy of these measurements in such topographically variable terrain is not well understood. For this study we measured 179 mountain summits with differential GNSS static surveys and compared summit elevation and location measurements to those measured by LiDAR in point cloud data sets. We measured summits in 13 US states (Washington, Idaho, Montana, Utah, California, Nevada, Arizona, New Mexico, Michigan, Wisconsin, Kentucky, Colorado, and Pennsylvania) and two Canadian provinces (British Columbia and Nova Scotia). Summits included icecapped peaks, open rocky peaks, and tree-covered peaks ranging in elevation from 490m to over 4000m. LiDAR-point-cloud-derived summit elevations and locations were computed using four different methods: manual processing, highest ground return, highest return, and Lastools reclassification. The average one-sigma LiDAR vertical errors for each method were 0.50m, 1.09m, 9.83m, and 1.96m, respectively. Average one-sigma horizontal errors were 3.03m, 2.41m, 5.17m, 3.78m, respectively. Errors are also presented separat

Predicting injuries and fatalities in traffic crashes plays a critical role in enhancing road safety, improving emergency response, and guiding public health interventions. This study investigates the added value of unstructured crash narratives (written by police officers at the scene) when combined with structured crash data to predict injury severity. Two widely used Natural Language Processing (NLP) techniques, Term Frequency-Inverse Document Frequency (TF-IDF) and Word2Vec, were employed to extract semantic meaning from the narratives, and their effectiveness was compared. To address the challenge of class imbalance, a K-Nearest Neighbors-based oversampling method was applied to the training data prior to modeling. The dataset consists of crash records from Kentucky spanning 2019 to 2023. To account for roadway heterogeneity, three road classification schemes were used: (1) eight detailed functional classes (e.g., Urban Two-Lane, Rural Interstate, Urban Multilane Divided), (2) four broader paired categories (e.g., Urban vs. Rural, Freeway vs. Non-Freeway), and (3) a unified dataset without classification. A total of 102 machine learning models were developed by combining struc

The field of environmental epidemiology has placed an increasing emphasis on understanding the health effects of mixtures of metals, chemicals, and pollutants in recent years. Bayesian Kernel Machine Regression (BKMR) is a statistical method that has gained significant traction in environmental mixture studies due to its ability to account for complex non-linear relationships between the exposures and health outcome and its ability to identify interaction effects between the exposures. However, BKMR makes the crucial assumption that the error terms have a constant variance, and this assumption is not typically checked in practice. In this paper, we create a diagnostic function for checking this constant variance assumption in practice and develop Heteroscedastic BKMR (HBKMR) for environmental mixture analyses where this assumption is not met. By specifying a Bayesian hierarchical variance model for the error term variance parameters, HBKMR produces updated estimates of the environmental mixture's health effects and their corresponding 95% credible intervals. We apply HBKMR in two real-world case studies that motivated this work: 1) Examining the effects of prenatal metal exposures

This paper introduces a user-friendly platform developed by the University of Kentucky Center for Applied AI, designed to make large, customized language models (LLMs) more accessible. By capitalizing on recent advancements in multi-LoRA inference, the system efficiently accommodates custom adapters for a diverse range of users and projects. The paper outlines the system's architecture and key features, encompassing dataset curation, model training, secure inference, and text-based feature extraction. We illustrate the establishment of a tenant-aware computational network using agent-based methods, securely utilizing islands of isolated resources as a unified system. The platform strives to deliver secure LLM services, emphasizing process and data isolation, end-to-end encryption, and role-based resource authentication. This contribution aligns with the overarching goal of enabling simplified access to cutting-edge AI models and technology in support of scientific discovery.

Accurate and timely regional weather prediction is vital for sectors dependent on weather-related decisions. Traditional prediction methods, based on atmospheric equations, often struggle with coarse temporal resolutions and inaccuracies. This paper presents a novel machine learning (ML) model, called MiMa (short for Micro-Macro), that integrates both near-surface observational data from Kentucky Mesonet stations (collected every five minutes, known as Micro data) and hourly atmospheric numerical outputs (termed as Macro data) for fine-resolution weather forecasting. The MiMa model employs an encoder-decoder transformer structure, with two encoders for processing multivariate data from both datasets and a decoder for forecasting weather variables over short time horizons. Each instance of the MiMa model, called a modelet, predicts the values of a specific weather parameter at an individual Mesonet station. The approach is extended with Re-MiMa modelets, which are designed to predict weather variables at ungauged locations by training on multivariate data from a few representative stations in a region, tagged with their elevations. Re-MiMa (short for Regional-MiMa) can provide highl

Thermal Desktop (TD) is an industry-standard thermal analysis tool used to create and analyze thermal models for landers, rovers, spacecraft, and instrument payloads. Currently, limited software exists to extract and visualize metrics relevant to heat flow within TD, impeding thermal engineers from analyzing their results quickly. This paper discusses a graphical user interface (GUI) built in MATLAB and C++ which uses TDs application programming interface (API), OpenTD, and a custom parser to address this void. Specifically, we present a method for efficiently loading temperature, conductance, and submodel metrics using a side effect of TDs Compressed Solution Results (CSR) files. This approach can reduce the runtime for correlating model nodes and conductors with submodel IDs by orders of magnitude. Lastly, we reflect on the shortcomings of this method for reading data, consider the future of the GUI, and provide recommendations for subsequent OpenTD releases.

We present efforts in the fields of machine learning and time series forecasting to accurately predict counts of future suspected opioid overdoses recorded by Emergency Medical Services (EMS) in the state of Kentucky. Forecasts help government agencies properly prepare and distribute resources related to opioid overdoses. Our approach uses county and district level aggregations of suspected opioid overdose encounters and forecasts future counts for different time intervals. Models with different levels of complexity were evaluated to minimize forecasting error. A variety of additional covariates relevant to opioid overdoses and public health were tested to determine their impact on model performance. Our evaluation shows that useful predictions can be generated with limited error for different types of regions, and high performance can be achieved using commonly available covariates and relatively simple forecasting models.

Developing and enforcing study protocols is a foundational component of medical research. As study complexity for participant interactions increases, translating study protocols to supporting application code becomes challenging. A collaboration exists between the University of Kentucky and Arizona State University to determine the efficacy of time-restricted eating in improving metabolic risk among postmenopausal women. This study utilizes a graph-based approach to monitor and support adherence to a designated schedule, enabling the validation and step-wise audit of participants' statuses to derive dependable conclusions. A texting service, driven by a participant graph, automatically manages interactions and collects data. Participant data is then accessible to the research study team via a website, which enables viewing, management, and exportation. This paper presents a system for automatically managing participants in a time-restricted eating study that eliminates time-consuming interactions with participants.

Coronary artery disease (CAD), one of the leading causes of mortality worldwide, necessitates effective risk assessment strategies, with coronary artery calcium (CAC) scoring via computed tomography (CT) being a key method for prevention. Traditional methods, primarily based on UNET architectures implemented on pre-built models, face challenges like the scarcity of annotated CT scans containing CAC and imbalanced datasets, leading to reduced performance in segmentation and scoring tasks. In this study, we address these limitations by introducing DINO-LG, a novel label-guided extension of DINO (self-distillation with no labels) that incorporates targeted augmentation on annotated calcified regions during self-supervised pre-training. Our three-stage pipeline integrates Vision Transformer (ViT-Base/8) feature extraction via DINO-LG trained on 914 CT scans comprising 700 gated and 214 non-gated acquisitions, linear classification to identify calcified slices, and U-NET segmentation for CAC quantification and Agatston scoring. DINO-LG achieved 89% sensitivity and 90% specificity for detecting CAC-containing CT slices, compared to standard DINO's 79% sensitivity and 77% specificity, red

Metallic foams are crucial to many emerging applications, among them shielding against hypervelocity impacts caused by micrometeoroids and orbital debris. The variability of properties at feature-scale and mesoscale lengths originating from the foam's inherently random microstructure makes predictive models of their properties challenging. It also hinders the optimization of components fabricated with such foams, an especially serious problem for spacecraft design where the balance between cost and mass must also be balanced against the catastrophic results of component failure. To address this problem, we compute the critical transition length between the feature-scale, where mechanical properties are determined by individual features, and the mesoscale, where behavior is determined by ensembles of features. At the mesoscale, distributions of properties -- with respect to both expectation value and standard variability -- are consistent and predictable. The Kentucky Random Structure Toolkit (KRaSTk) is applied to determine the transition from feature-scale to mesoscale for computational volumes representing metallic foams at a range of reduced densities. The transition is found to

AI systems may be better thought of as peers than as tools. This paper explores applications of augmented collective intelligence (ACI) beneficial to collaborative ideation. Design considerations are offered for an experiment that evaluates the performance of hybrid human- AI collectives. The investigation described combines humans and large language models (LLMs) to ideate on increasingly complex topics. A promising real-time collection tool called Polis is examined to facilitate ACI, including case studies from citizen engagement projects in Taiwan and Bowling Green, Kentucky. The authors discuss three challenges to consider when designing an ACI experiment: topic selection, participant selection, and evaluation of results. The paper concludes that researchers should address these challenges to conduct empirical studies of ACI in collaborative ideation.

Although autonomy has gained widespread usage in structured and controlled environments, robotic autonomy in unknown and off-road terrain remains a difficult problem. Extreme, off-road, and unstructured environments such as undeveloped wilderness, caves, rubble, and other post-disaster sites pose unique and challenging problems for autonomous navigation. Based on our participation in the DARPA Subterranean Challenge, we propose an approach to improve autonomous traversal of robots in subterranean environments that are perceptually degraded and completely unknown through a traversability and planning framework called STEP (Stochastic Traversability Evaluation and Planning). We present 1) rapid uncertainty-aware mapping and traversability evaluation, 2) tail risk assessment using the Conditional Value-at-Risk (CVaR), 3) efficient risk and constraint-aware kinodynamic motion planning using sequential quadratic programming-based (SQP) model predictive control (MPC), 4) fast recovery behaviors to account for unexpected scenarios that may cause failure, and 5) risk-based gait adaptation for quadrupedal robots. We illustrate and validate extensive results from our experiments on wheeled a

Utilizing data available from the Kentucky Geonet (KYGeonet.ky.gov) the fossil fuel mining locations created by the Kentucky Geological Survey geo-locating oil and gas wells are mapped using ESRI ArcGIS in Kentucky single plain 1602 ft projection. This data was then exported into a spreadsheet showing latitude and longitude for each point to be used for modeling at different scales to determine the fractal dimension of the set. Following the porosity and diffusivity studies of Tarafdar and Roy1 we extract fractal dimensions of the fossil fuel mining locations and search for evidence of scaling laws for the set of deposits. The Levy index is used to determine a match to a statistical mechanically motivated generalized probability function for the wells. This probability distribution corresponds to a solution of a dynamical anomalous diffusion equation of fractional order that describes the Levy paths which can be solved in the diffusion limit by the Fox H function ansatz.