Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “training”

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.

At least 1,639 records · Page 91
All sourcesNASA NTRSDOE OSTI
DOE OSTI2011

Controlling the Spacing of Attosecond Pulse Trains from Relativistic Surface Plasmas

When a laser pulse hits a solid surface with relativistic intensities, XUV attosecond pulses are generated in the reflected light. We present an experimental and theoretical study of the temporal properties of attosecond pulse trains in this regime. The recorded harmonic spectra show distinct fine structures which can be explained by a varying temporal pulse spacing that can be controlled by the laser contrast. The pulse spacing is directly related to the cycle-averaged motion of the reflecting surface. Thus the harmonic spectrum contains information on the relativistic plasma dynamics.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2023

Stable Solid Molecular Hydrogen above 900 K from a Machine-Learned Potential Trained with Diffusion Quantum Monte Carlo

Here, we survey the phase diagram of high-pressure molecular hydrogen with path integral molecular dynamics using a machine-learned interatomic potential trained with quantum Monte Carlo forces and energies. Besides the HCP and C2/c–24 phases, we find two new stable phases both with molecular centers in the Fmmm–4 structure, separated by a molecular orientation transition with temperature. The high temperature isotropic Fmmm–4 phase has a reentrant melting line with a maximum at higher temperature (1450 K at 150 GPa) than previously estimated and crosses the liquid-liquid transition line around 1200 K and 200 GPa.

08 HYDROGEN↗
DOE OSTI2020

Dynamic pulse-to-pulse thermal load effects in pulse-train-mode self-seeded X-ray free-electron laser

Thermal load has been a haunting factor that undermines the brightness and coherence of the self-seeded X-ray free-electron laser. Additionally, different from uniformly pulsed mode, in pulse train mode a thermal quasi-steady state of the crystal monochromator may not be reached. This leads to a dynamic thermal distortion of the spectral transmission curves and seed quality degradation. In this paper, the pulse-to-pulse thermal load effects on the spectral transmission curves and seed quality are shown, and some instructive information for the tuning process is provided.

47 OTHER INSTRUMENTATION↗
DOE OSTI2020

Efficient I/O for Neural Network Training with Compressed Data

FanStore is a shared object store that enables efficient and scalable neural network training on supercomputers. By providing a global cache layer on node-local burst buffers using a compressed representation, it significantly enhances the processing capability of deep learning (DL) applications on existing hardware. In addition, FanStore allows POSIX-compliant file access to the compressed data in user space. We investigate the tradeoff between runtime overhead and data compression ratio using real-world datasets and applications, and propose a compressor selection algorithm to maximize storage capacity given performance constraints. We consider both asynchronous (i.e., with prefetching) and synchronous I/O strategies, and propose mechanisms for selecting compressors for both approaches. Using FanStore, the same storage hardware can host 2-13x more data for example applications without significant runtime overhead. Empirically, our experiments show that FanStore scales to 512 compute nodes with near linear performance scalability.

Zhang, Zhao↗
DOE OSTI2024

Adaptive Quantum Generative Training using an Unbounded Loss Function

We propose a generative quantum learning algorithm using the Adaptive Derivative-Assembled Problem Tailored ansatz (ADAPT) framework in which the loss function to be minimized is the maximal quantum Rényi divergence of order two, an unbounded function that mitigates barren plateaus which inhibit training variational circuits. We benchmark this method against other state-of-the-art adaptive algorithms by learning random two-local thermal states. We perform numerical experiments of up to 12 qubits comparing our method learning algorithms that use linear objective functions and show that Rényi-ADAPT is capable of constructing shallow quantum circuits competitive with existing methods, while the gradients remain favorable resulting from the maximal Rényi divergence loss function.

quantum algorithms, quantum machine learning, quan↗
DOE OSTI2024

An Efficient Checkpointing System for Large Machine Learning Model Training

As machine learning models increase in size and complexity rapidly, the cost of checkpointing in ML training became a bottleneck in storage and performance (time). For example, the latest GPT-4 model has massive parameters at the scale of 1.76 trillion. It is highly time and storage consuming to frequently writes the model to checkpoints with more than 1 trillion floating point values to storage. This work aims to understand and attempt to mitigate this problem. First, we characterize the checkpointing interface in a collection of representative large machine learning/language models with respect to storage consumption and performance overhead. Second, we propose the two optimizations: i) A periodic cleaning strategy that periodically cleans up outdated checkpoints to reduce the storage burden; ii) A data staging optimization that coordinates checkpoints between local and shared file systems for performance improvement.

machine learning, artificial intelligence↗
DOE OSTI2024

Assessment of Training Performance, Degradation and Robustness of Paraffin-Wax Impregnated Nb 3 Sn Demonstrator Under High Magnetic Field

In the context of high-energy physics, the use of Nb 3 Sn superconducting magnets as a cost-effective and reliable technology depends on improvements in the following areas: long development and manufacturing cycles, conductor degradation after thermal cycling, long training, as well as a demonstration in accelerator magnets with a beam aperture of the full potential of modern Nb 3 Sn conductors. In short, performance, robustness, and cost are the three issues to be addressed. The Magnet Development project (MagDev) of the Swiss Accelerator Research and Technology initiative (CHART) at the Paul Scherrer Institute (PSI) aims to contribute to the solutions to each of these issues, re-thinking the manufacturing and design process. Here in our program, every innovation is to be validated by means of a panoply of fast-turnaround tools: from non-powered and powered samples and coils, tested under background field, to low-field subscale magnets and high field short prototypes. This work presents one element in this panoply of R&D vehicles: a stress-managed Nb 3 Sn coil called BigBOX, impregnated with paraffin wax, and tested, through a collaboration with the Magnet Development Program of the United States (US-MDP), in the background field of Brookhaven National Laboratory (BNL)’s common coils dipole DCC17.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗
DOE OSTI2018

Sparse Coding for N-Gram Feature Extraction and Training for File Fragment Classification

File fragment classification is an important step in the task of file carving in digital forensics. In file carving, files must be reconstructed based on their content as a result of their fragmented storage on disk or in memory. Existing methods for classification of file fragments typically use hand-engineered features such as byte histograms or entropy measures. In this paper, we propose an approach using sparse coding that enables automated feature extraction. Sparse coding, or sparse dictionary learning, is an unsupervised learning algorithm, and is capable of extracting features based simply on how well those features can be used to reconstruct the original data. With respect to file fragments, we learn sparse dictionaries for n-grams, continuous sequences of bytes, of different sizes. These dictionaries may then be used to estimate n-gram frequencies for a given file fragment, but for significantly larger n-gram sizes than are typically found in existing methods which suffer from combinatorial explosion. To demonstrate the capability of our sparse coding approach, we used the resulting features to train standard classifiers such as support vector machines (SVMs) over multiple file types. Experimentally, we achieved significantly better classification results with respect to existing methods, especially when the features were used in supplement to existing hand-engineered features.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2015

SU-E-I-62: Reduction of Susceptibility Artifacts by Increasing the Bandwidth (BW) and Echo Train Length (ETL)

Purpose: The aim of this present study is to increase bandwidth (BW) and echo train length (ETL) in Proton Density Turbo Spin Echo (PD TSE) sequences with and without fat saturation (FS) as well as in Turbo Inversion Recovery Magnitude sequences (TIRM) in order to assess whether these sequences are capable of reducing susceptibility artifacts. Methods: We compared 1) TIRM coronal (COR) with the same sequence with increased both BW and ETL 2) Conventional PD TSE sagittal (SAG) with FS with an increased BW 3) Conventional PD TSE SAG without FS with an increased BW 4) Conventional PD TSE SAG without FS with increased both BW and ETL. A quantitative analysis was performed to measure the extent of the susceptibility artifacts. Furthermore, a qualitative analysis was performed by two radiologists in order to evaluate the susceptibility artifacts, image distortion and fat suppression. The depiction of cartilage, menisci, muscles, tendons and bone marrow were also qualitatively analyzed. Results: The quantitative analysis found that the modified TIRM sequence is significantly superior to the conventional one regarding the extent of the susceptibility artifacts. In the qualitative analysis, the modified TIRM sequence was superior to the corresponding conventional one in eight characteristics out of ten that were analyzed. The modified PD TSE with FS was superior to the corresponding conventional one regarding the susceptibility artifacts, image distortion and depiction of bone marrow and cartilage while achieving effective fat saturation. The modified PD TSE sequence without FS with a high (H) BW was found to be superior corresponding to the conventional one in the case of cartilage. Conclusion: Consequently, TIRM sequence with an increased BW and ETL is proposed for producing images of high quality and modified PD TSE with H BW for smaller metals, especially when FS is used.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Quantitative cross-species translators of cardiac myocyte electrophysiology: Model training, experimental validation, and applications

Animal experimentation is key in the evaluation of cardiac efficacy and safety of novel therapeutic compounds. However, interspecies differences in the mechanisms regulating excitation-contraction coupling can limit the translation of experimental findings from animal models to human physiology and undermine the assessment of drugs’ efficacy and safety. Here, we built a suite of translators for quantitatively mapping electrophysiological responses in ventricular myocytes across species. We trained these statistical operators using a broad dataset obtained by simulating populations of our biophysically detailed computational models of action potential and Ca 2+ transient in mouse, rabbit, and human. We then tested our translators against experimental data describing the response to stimuli, such as ion channel block, change in beating rate, and β-adrenergic challenge. We demonstrate that this approach is well suited to predicting the effects of perturbations across different species or experimental conditions and suggest its integration into mechanistic studies and drug development pipelines.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2021

Mixture density network estimation of continuous variable maximum likelihood using discrete training samples

Abstract Mixture density networks (MDNs) can be used to generate posterior density functions of model parameters $$\varvec{\theta }$$ θ given a set of observables $${\mathbf {x}}$$ x . In some applications, training data are available only for discrete values of a continuous parameter $$\varvec{\theta }$$ θ . In such situations, a number of performance-limiting issues arise which can result in biased estimates. We demonstrate the usage of MDNs for parameter estimation, discuss the origins of the biases, and propose a corrective method for each issue.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2022

Distributed Training and Optimization of Neural Networks

Deep learning models are yielding increasingly better performances thanks to multiple factors. To be successful, model may have large number of parameters or complex architectures and be trained on large dataset. This leads to large requirements on computing resource and turn around time, even more so when hyperparameter optimization is done (e.g. search over model architectures). While this is a challenge that goes beyond particle physics, we review the various ways to do the necessary computations in parallel, and put it in the context of high-energy physics.

Vlimant, Jean-Roch↗
DOE OSTI2018

Integrated Model, Batch and Domain Parallelism in Training Neural Networks

We propose a new integrated method of exploiting model, batch and domain parallelism for the training of deep neural networks (DNNs) on large distributed-memory computers using minibatch stochastic gradient descent (SGD). Our goal is to find an efficient parallelization strategy for a fixed batch size using $P$ processes. Our method is inspired by the communication-avoiding algorithms in numerical linear algebra. We see $P$ processes as logically divided into a $P_r \times P_c$ grid where the $P_r$ dimension is implicitly responsible for model/domain parallelism and the $P_c$ dimension is implicitly responsible for batch parallelism. In practice, the integrated matrix-based parallel algorithm encapsulates these types of parallelism automatically. We analyze the communication complexity and analytically demonstrate that the lowest communication costs are often achieved neither with pure model nor with pure data parallelism. We also show how the domain parallel approach can help in extending the theoretical scaling limit of the typical batch parallel method.

97 MATHEMATICS AND COMPUTING↗