InterAug: A Tuning-Free Augmentation Policy for Data-Efficient and Robust Object Detection
Explore the source record for details and available documents.
SEARCH · Engineering Papers
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Not provided.
Not Available
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Not provided.
Explore the source record for details and available documents.
Inspired by the connectivity mechanisms in the brain, neuromorphic computing architectures model Spiking Neural Networks (SNNs) in silicon. As such, neuromorphic architectures are designed and developed with the goal of having small, low power chips that can perform control and machine learning tasks. However, the power consumption of the developed hardware can greatly depend on the size of the network that is being evaluated on the chip. Furthermore, the accuracy of a trained SNN that is evaluated on chip can change due to voltage and current variations in the hardware that perturb the learned weights of the network. While efforts are made on the hardware side to minimize those perturbations, a software based strategy to make the deployed networks more resilient can help further alleviate that issue. In this work, we study Spiking Neural Networks in two neuromorphic architecture implementations with the goal of decreasing their size, while at the same time increasing their resiliency to hardware faults. We leverage an evolutionary algorithm to train the SNNs and propose a multiobjective fitness function to optimize the size and resiliency of the SNN. We demonstrate that this strategy leads to well-performing, small-sized networks that are more resilient to hardware faults.
Explore the source record for details and available documents.
It is commonly accepted that the introduction of hydrogen as an energy carrier for light-duty vehicles involves concomitant technological development of infrastructure elements, such as production, delivery, and consumption, all associated with certain emission levels. To analyze these at a system level, the suite of corresponding models developed by the United States Department of Energy andinvolving several national laboratories is combined in one macro-system model (MSM). The macro-system model is being developed as a cross-cutting analysis tool that combines a set of hydrogen technology analysis models. Within the MSM, a federated simulation framework is used for consistent data transfer between the component models. The framework is built to suit cross-model as well ascross-platform data exchange and involves features of 'over-the-net' computation.
This paper reports a verification study for a method that fits functions to sets of data from several experiments simultaneously. The method finds a maximum a posteriori probability (MAP) estimate of a function subject to constraints (e. g., convexity in the study), uncertainty about the estimate, and a quantitative characterization of how data from each experiment constrains that uncertainty. While the present work focuses on a model of the Equation Of State (EOS) of gasses produced by detonating a high explosive, the method can be applied to a wide range of physics processes with either parametric or semi-parametric models. As a verification exercise, a reference EOS is used and artificial experimental data sets are created using numerical integration of ordinary differential equations and pseudo-random noise. The method yields an estimate of the EOS that is close to the reference, and identifies how each experiment most constrains the result.