136879 research outputs found
Sort by
Robust Distributed Learning of Functional Data From Simulators through Data Sketching
No full textRealistic simulations are crucial for comprehending complex systems in climate and environmental studies. Yet, running sophisticated computational models across a wide range of input settings often overwhelms large computer systems. Statistical surrogate models, or emulators, play a vital role in efficiently exploring the simulator input space. Functional data models involving Gaussian processes (GPs) and their computationally efficient variants have become standard tools for achieving this goal. The conventional centralized processing of such models requires substantial computational and storage resources at the central server. To counter this, emerging distributed Bayesian learning frameworks partition raw data into shards and distribute computations of these shards across machines. While this strategy mitigates data storage costs and improves computation within each machine, concerns arise regarding the sensitivity of distributed inference to shard selection.
Motivated by the concept of data sketching in the literature, this article proposes an innovative alternative. Instead of creating data shards, our approach employs multiple random matrices to construct multiple random linear projections, or `"random sketches," of the complete dataset. Posterior inference on functional data models is performed using random sketches on various machines in parallel. These individual inferences are then combined across machines at a central server. By aggregating inference across diverse random matrices, our approach proves resilient to the selection of data sketches, leading to the development of novel robust distributed Bayesian learning approach.
An important advantage of our approach is its ability to maintain the privacy of sampling units, as the inference is based on random data sketches that do not allow the recovery of raw data.
We illustrate the significance of our approach through various simulated data examples in the realm of Bayesian distributed learning techniques. Finally, we demonstrate the performance of our proposed approach as an emulator with surrogates of the Sea, Lake, and Overland Surges from Hurricanes (SLOSH) simulator���a choice of simulator for government agencies.National Science Foundation, Los Alamos National Laboratorie
Understanding the Molecular Role and Mechanisms of SmokTcr in the Mouse t-Haplotype
Get PDFThrough predation, disease transmission, and increased competition, invasive species are one of the major agents of global biodiversity loss. Current eradication methods, such as the spread of anticoagulant rodent baits, pose several disadvantages: harmful secondary effects on native species and humans, negative public perception, and not feasible in densely populated and geographically complex areas. Identified as an alternative eradication method, gene drives can be manipulated to develop species-specific genetic biocontrol tools. The t-haplotype is a naturally existing murine gene drive that confers biased inheritance of the t-haplotype over the wild-type allele in the progeny of t-heterozygous males. This phenomenon is called transmission ratio distortion (TRD). The current model of TRD suggests that, in t-heterozygous males, t-carrying sperm express distorters that disrupt the motility of both wild type sperm and t-sperm. Simultaneously, however, t-sperm also expresses a responder that restores its own motility. Sperm motility kinase SmokTcr (Tcr)���a hybrid gene formed through the fusion of Smok and Rps6ka2 (commonly referred to as Rsk3)���has been proposed to be the responder that rescues t-sperm. To determine if Tcr is the sole responder protein that is essential for TRD to occur, our project aimed to assess tw2 transmission rates in the absence of Tcr. CRISPR-Cas9 will be used to knockout Rsk3 and inhibit the formation of Tcr. We hypothesize that if Tcr is absent, then the distorted ~95% inheritance ratio of the t-haplotype will not be observed in the progeny of tw2-heterozygous males. Furthermore, tw2-heterozygous males may be sterile since the rescue mechanism for t-sperm will be disrupted. The findings from this study will aid scientists in designing genetic biocontrol tools to potentially employ against invasive mice populations
Efficacy of Streptokinase as a Thrombolytic Agent Against Plasma-Only In Vitro Thrombi
Get PDFBlood clots, also known as thrombi, can be life-threatening medical emergencies. Blood-contacting medical devices have a tendency to induce thrombus formation, which can result in the failure of the device or other complications associated with thrombosis such as heart attacks and strokes. Additional methods are needed for creating in vitro thrombi that facilitate efforts to evaluate the efficacy of various thrombolytic therapies in instances of medical device-induced thrombosis. Calcium chloride was added to plasma to induce clotting within a rotating mechanical apparatus. Thrombi were then allowed to coagulate for various lengths of time before streptokinase, a thrombolytic drug selected for its practicality and history of success in instances of medical device-induced thrombosis, was administered within the apparatus. The degree to which the thrombus was dissolved, as well as the length of time this took, were recorded. Findings regarding the correlation between the age of a thrombus and the efficacy of streptokinase as a thrombolytic agent for plasma-only clots created in this manner were compared to the results of previous studies. This proof of concept experiment suggests that the rotating mechanical apparatus is an effective method for creating in vitro plasma-only clots, and for testing the efficacy of thrombolytic drugs in such an environment. Further exploration and development of this in vitro method using whole-blood clots of various origins and a variety of thrombolytic therapies has the potential to make progress toward standardizing thrombolytic protocols in instances of medical device-induced thrombosis
2017 Texas High Plains and Panhandle Replicated Agronomic Cotton Evaluation (RACE) Trials
No full text
Local Off-Road Navigation
Get PDFThis work describes a local mapping and planning software stack for off-road autonomous ground vehicles. We start with a local 3D voxel mapping framework for off-road path planning and navigation. This method provides both hard and soft positive obstacle detection, negative obstacle detection, slope estimation, and roughness estimation. The system can provide online performance by using a 3D array lookup table data structure and leveraging the GPU. This system is capable of making use of the more traditional lidar. However, the limitations of lidar, such
as reduced performance in harsh environmental conditions and limited range, have prompted the exploration of alternative sensing technologies. We, therefore, investigate the potential of radar for off-road local navigation, as it offers the advantages of a longer range and the ability to penetrate dust and light vegetation. By adapting our lidar-based voxel methods for radar, we evaluate its performance against lidar under various off-road conditions. We show that radar can provide a
significant range advantage over lidar while maintaining accuracy for ground plane estimation and obstacle detection. Once lidar and radar data have been accumulated into the local map, we combine obstacle detection and a terrain gradient map to make a simple and adaptable cost map.
However, due to the hand-tuned nature of this layered approach, we also explore the potential of using a physics-based cost function. This physics-based function uses terrain roughness information and a model of the vehicle���s dynamics to create a maximum speed map. This map generates a costmap to solve for minimum-time local paths. Finally, once the cost map is created, we generate the local path for the vehicle to follow. An A* planner finds optimal paths over this map. We then take multiple samples over the control input space and do a kinematic forward simulation to generate feasible trajectories. Then, the most optimal trajectory, as determined by the cost map and proximity to the A* path, is chosen. However, these resulting paths do not necessarily consider all of the vehicle���s kinematic constraints. Therefore, we also proposed a path optimization framework considering these kinematic constraints. By thinking in the actuator space, we can represent such
constraints as limits in the space rather than derived properties of the path. We present an actuator space approach to path optimization for off-road ground vehicles. This optimization is done by representing the path as a list of steering angles over the path length, transforming the set of kinematic constraints into constraints on the steering angle. We then put this path into a gradient descent solver, producing kinematically feasible and optimized paths per our cost function. Once this optimization step is completed, the final path is then sent to the vehicle. Finally, the effectiveness of our methods is proven by multiple tests over off-road terrain on both real and simulated vehicles. We achieved a fully autonomous speed of 5 m/s with the entire system.
Additionally, we demonstrate successful autonomous navigation at a speed of 2.5 m/s over a path length of 350 m using only radar for ground plane estimation and obstacle detection. We believe this is the first time this has been done using radar alone. Finally, we show that our mapping system is capable of high-speed operation on multiple vehicles at speeds of 5 m/s in autonomous operation and 12 m/s in manual operation with a map update rate of 10 Hz