100039 research outputs found
Sort by
Exploring Runtime Sparsification of YOLO Model Weights During Inference
No full textIn the pursuit of real-time object detection with constrained computational resources, the optimization of neural network architectures is paramount. We introduce novel sparsity induction methods within the YOLOv4-Tiny framework to significantly improve computational efficiency while maintaining high accuracy in pedestrian detection. We present three sparsification approaches: Homogeneous, Progressive, and Layer-Adaptive, each methodically reducing the model’s complexity without compromising its detection capability. Additionally, we refine the model’s output with a memory-efficient sliding window approach and a Bounding Box Sorting Algorithm, ensuring precise Intersection over Union (IoU) calculations. Our results demonstrate a substantial reduction in computational load by zeroing out over 50% of the weights with only a minimal 6% loss in IoU and 0.6% loss in F1-Score
Raw data files contributing to Changes in ApoE and TIMP-1 Expression Correlate with Outer Blood-Retinal Barrier Disruption in an In Vitro Model of Retinal Aging
No full textAge-related macular degeneration (AMD) is a leading cause of blindness worldwide. Unfortunately, the early stages of this disease are poorly understood, which has led to limited treatment options. Investigating normal changes in tissues eventually affected by AMD can further elucidate the mechanisms of disease progression and lead to novel therapeutic targets. The primary cell layer affected in AMD is the retinal pigment epithelium (RPE), which forms the outer blood-retinal barrier (oBRB). Beneath the RPE lies Bruch’s membrane, a proteinaceous layer that naturally thickens and stiffens with age. These changes to Bruch’s membrane are also implicated in RPE dysfunction and AMD progression. To investigate the relationship between normal, age-related changes in Bruch’s membrane and AMD development, we engineered a tunable in vitro model of Bruch’s membrane to support primary porcine RPE cells. We performed transepithelial electrical resistance (TEER) measurements, viability assays, morphological analysis, immunocytochemistry, and enzyme-linked immunosorbent assays (ELISA) to evaluate monolayer integrity and angiogenic factor expression. Cells cultured on our aged model exhibited changes similar to those seen in AMD, including reduced monolayer integrity, the formation of sub-RPE deposits, and eventual cell death. Notably, apolipoprotein E (ApoE), a known drusen component and Alzheimer’s disease marker, was overexpressed prior to deposit accumulation and cell death. Regions of ApoE overexpression corresponded with disrupted expression of zonula occludens-1, a junctional protein. While most angiogenic factors remained unchanged, tissue inhibitor of metalloproteinases-1 (TIMP-1) was transiently overexpressed before cell death. These findings suggest that ApoE and TIMP-1 may play key roles in early AMD pathogenesis and represent potential targets for future therapeutic interventio
Training on Learn the Signs. Act Early. in Head Start and Early Head Start Shows Ineffective Without Supporting Implementation Policy
No full textEarly identification and intervention for children with developmental delays improves child outcomes, yet many children are not monitored, screened, or identified early despite its effectiveness. The relationship between the use of the Centers for Disease Control and Prevention’s “Learn the Signs. Act Early.” developmental monitoring program and referrals to intervention services is not well understood. This study investigated how “Learn the Signs. Act Early.” implementation practices within Massachusetts’ Head Start and Early Head Start agencies correlated with the rate of referrals to intervention services. Researchers utilized a non-experimental quantitative design. Secondary data on referral rates from Head Start and Early Head Start agencies, combined with a cross-sectional survey to understand developmental monitoring practices within Head Start and Early Head Start programs, was analyzed. Results indicated that developmental monitoring training was not significantly correlated with referral rates to intervention services. Despite high familiarity, surveyed agencies showed varying levels of “Learn the Signs. Act Early.” material implementation, training, and policy use. These findings indicate the need for universal developmental monitoring and screening policies for Head Start and Early Head Start programs to help improve access to critical early developmental services for young children
How a 2 to 5-Year Experimental Lake Powell and Lake Mead Release Program Tied to Reservoir Inflows can be a Win for Adaptive Risk Management
No full textLake Powell and Lake Mead are at risk of drawdown to their minimum power and dead pools in the next few years because current and proposed shortage and release operations tied to reservoir storage and sometimes prior natural flow cannot keep pace with U.S. Bureau of Reclamation’s numerous scenarios of more volatile, declining, and longer-lasting periods of low flows. One experimental program to reduce risk can instead adapt reservoir releases to monitored changes in physical reservoir inflow and reservoir evaporation. First, stabilize reservoir storage by temporarily setting reservoir release to the physical reservoir inflow minus evaporation (the available water). Second, continue to stabilize storage by changing releases to match changes in physical reservoir inflow and evaporation. Third, build storage by decreasing releases from the release needed to stabilize reservoir storage. An experimental program has additional wins such as it can begin immediately or at any target reservoir elevation without the need for new agreements. A program can also stabilize and build reservoir storage even when low flows persist. Users who hold back some of their share of reservoir releases for later release can customize and adapt their strategies to manage future risks of water shortages. An experimental risk management program can help build operational experience and flexibility in a new era of low reservoir storage, volatile, declining, and longer-lasting low flows. We share links to further explore some of our new risk communication and adaptive management tools
FlaPLeT: A Full-Stack Web Platform for End-To-End Time Series Data Processing and Machine Learning in Solar Flare Prediction
No full textSolar flare prediction is a central challenge in space weather forecasting, with direct implications for satellite operations, aviation safety, and power grid reliability. Machine learning has achieved state-of-the-art performance for this task, particularly when applied to photospheric magnetic field parameters. FlaPLeT is an open-source, full-stack web platform that supports end-to-end machine learning workflows for multivariate time-series–based solar flare prediction without requiring any coding expertise. Built with React, Django, Celery, and PostgreSQL, the system integrates dataset preprocessing, data augmentation, functional network (graph) construction, and machine learning model training into modular asynchronous tasks that generate downloadable datasets, trained models, and structured JSON reports. The platform is deployed on a dedicated Windows server using NGINX, Waitress, Redis, TLS encryption, and reCAPTCHA to ensure secure and scalable operation. FlaPLeT lowers the barrier for heliophysicists to apply machine learning to photospheric magnetic field data and to systematically evaluate how preprocessing strategies and hyperparameter choices affect flare-prediction accuracy. Its cloud-based deployment removes local hardware constraints and makes the platform accessible to researchers worldwide through a standard web browser
Sunflower Cut Flower Production in Utah
No full textSunflower is low-maintenance, full-sun crop that is well-adapted to Utah’s hot and dry conditions. Single-stem cultivars are most popular for cut flower production. This fact sheet provides information on sunflower production, including cultivars, site preparation, germination, irrigation, pests, and disease, nutrient management, harvest and storage, postharvest bed management, and economics
Academic Standards Subcommittee Agenda January 15, 2026
No full textWelcome and approval of December 18, 2025 minutes Discussion Topics: Next Meeting: Thursday, February 19, 202
Comment on the Draft Environmental Impact Statement for New Operations for Lake Powell and Lake Mead Post-2026
No full textWe appreciate the Bureau of Reclamation’s extensive effort in preparing the Post-2026 Draft Environmental Impact Statement (Draft EIS) for operational guidelines for Lake Powell, Lake Mead, and the Colorado River Basin. This commentary will address positives, questions, and concerns from the Draft EIS. We also share recommendations for consideration in the Final EIS
Faculty Senate Executive Committee Minutes January 20, 2026
No full textCall to Order University Business Faculty Senate Business Reports School of Graduate Studies Professional Responsibilities and Procedures Committee Office of Research Old Business New Business Information Adjourn: 4:30 P
Black Medic (Medicago lupulina) Identification and Control in Residential Landscapes
No full textBlack medic (Medicago lupulina) is an herbaceous winter or summer annual broadleaf landscape weed. This fact sheet provides a weed description and describes suppression and control options