-
Feasibility Study on Active Learning of Smart Surrogates for Scientific Simulations
Authors:
Pradeep Bajracharya,
Javier Quetzalcóatl Toledo-Marín,
Geoffrey Fox,
Shantenu Jha,
Linwei Wang
Abstract:
High-performance scientific simulations, important for comprehension of complex systems, encounter computational challenges especially when exploring extensive parameter spaces. There has been an increasing interest in develo** deep neural networks (DNNs) as surrogate models capable of accelerating the simulations. However, existing approaches for training these DNN surrogates rely on extensive…
▽ More
High-performance scientific simulations, important for comprehension of complex systems, encounter computational challenges especially when exploring extensive parameter spaces. There has been an increasing interest in develo** deep neural networks (DNNs) as surrogate models capable of accelerating the simulations. However, existing approaches for training these DNN surrogates rely on extensive simulation data which are heuristically selected and generated with expensive computation -- a challenge under-explored in the literature. In this paper, we investigate the potential of incorporating active learning into DNN surrogate training. This allows intelligent and objective selection of training simulations, reducing the need to generate extensive simulation data as well as the dependency of the performance of DNN surrogates on pre-defined training simulations. In the problem context of constructing DNN surrogates for diffusion equations with sources, we examine the efficacy of diversity- and uncertainty-based strategies for selecting training simulations, considering two different DNN architecture. The results set the groundwork for develo** the high-performance computing infrastructure for Smart Surrogates that supports on-the-fly generation of simulation data steered by active learning strategies to potentially improve the efficiency of scientific simulations.
△ Less
Submitted 12 July, 2024; v1 submitted 10 July, 2024;
originally announced July 2024.
-
Light and microwave driven spin pum** across FeGaB-BiSb interface
Authors:
Vinay Sharma,
Weipeng Wu,
Prabesh Bajracharya,
Duy Quang To,
Anthony Johnson,
Anderson Janotti,
Garnett W. Bryant,
Lars Gundlach,
M. Benjamin Jungfleisch,
Ramesh C. Budhani
Abstract:
3-D topological insulators (TI) with large spin Hall conductivity have emerged as potential candidates for spintronic applications. Here, we report spin to charge conversion in bilayers of amorphous ferromagnet (FM) Fe_{78}Ga_{13}B_{9} (FeGaB) and 3-D TI Bi_{85}Sb_{15} (BiSb) activated by two complementary techniques: spin pum** and ultrafast spin-current injection. DC magnetization measurements…
▽ More
3-D topological insulators (TI) with large spin Hall conductivity have emerged as potential candidates for spintronic applications. Here, we report spin to charge conversion in bilayers of amorphous ferromagnet (FM) Fe_{78}Ga_{13}B_{9} (FeGaB) and 3-D TI Bi_{85}Sb_{15} (BiSb) activated by two complementary techniques: spin pum** and ultrafast spin-current injection. DC magnetization measurements establish the soft magnetic character of FeGaB films, which remains unaltered in the heterostructures of FeGaB-BiSb. Broadband ferromagnetic resonance (FMR) studies reveal enhanced dam** of precessing magnetization and large value of spin mixing conductance (5.03 x 10^{19} m^{-2}) as the spin angular momentum leaks into the TI layer. Magnetic field controlled bipolar dc voltage generated across the TI layer by inverse spin Hall effect is analyzed to extract the values of spin Hall angle and spin diffusion length of BiSb. The spin pum** parameters derived from the measurements of the femtosecond light-pulse-induced terahertz emission are consistent with the result of FMR. Kubo-Bastin formula and tight-binding model calculations shed light on the thickness-dependent spin-Hall conductivity of the TI films, with predictions that are in remarkable agreement with the experimental data. Our results suggest that room temperature deposited amorphous and polycrystalline heterostructures provide a promising platform for creating novel spin orbit torque devices.
△ Less
Submitted 12 January, 2022;
originally announced January 2022.
-
Resonant Precession of Magnetization and Precession -- Induced DC voltages in FeGaB Thin Films
Authors:
Prabesh Bajracharya,
Vinay Sharma,
Anthony Johnson,
Ramesh C. Budhani
Abstract:
Measurements of frequency dependent ferromagnetic resonance (FMR) and spin pum** driven dc voltage (V_{dc}) are reported for amorphous films of Fe_{78}Ga_{13}B_{9} (FeGaB) alloy to address the phenomenon of self-induced inverse spin Hall effect (ISHE) in plain films of metallic ferromagnets. The V_{dc} signal, which is antisymmetric on field reversal, comprises of symmetric and asymmetric Lorent…
▽ More
Measurements of frequency dependent ferromagnetic resonance (FMR) and spin pum** driven dc voltage (V_{dc}) are reported for amorphous films of Fe_{78}Ga_{13}B_{9} (FeGaB) alloy to address the phenomenon of self-induced inverse spin Hall effect (ISHE) in plain films of metallic ferromagnets. The V_{dc} signal, which is antisymmetric on field reversal, comprises of symmetric and asymmetric Lorentzians centered around the resonance field. Dominant role of thin film size effects is seen in setting the magnitude of static magnetization, V_{dc} and dynamics of magnetization precession in thinner films (\leq 8 nm). The film thickness dependence of magnetization parameters indicates the presence of a magnetically disordered region at the film-substrate interface, which may promote preferential flow of spins generated by the precessing magnetization towards the substrate. However, the V_{dc} signal also draws contributions from rectification effects of a \approx 0.4 \% anisotropic magnetoresistance and a large (\approx 54 nΩ.m) anomalous Hall resistivity (AHR) of these films which ride over the effect of spin-orbit coupling driven spin-to-charge conversion near the film-substrate interface. We have addressed these data in the framework of the existing theories of electrodynamics of a ferromagnetic film subjected to radio-frequency field in a coplanar waveguide geometry. Our estimation of the self-induced ISHE for the sample with 54 nΩ.m AHR shows that it may contribute significantly (\approx 90\%) to the measured symmetric voltage. This study is expected to be very useful for fully understanding the spin pum** induced dc voltages in metallic ferromagnets with disordered interfaces and large anomalous Hall effect.
△ Less
Submitted 10 January, 2022;
originally announced January 2022.
-
Fast Posterior Estimation of Cardiac Electrophysiological Model Parameters via Bayesian Active Learning
Authors:
Md Shakil Zaman,
Jwala Dhamala,
Pradeep Bajracharya,
John L. Sapp,
B. Milan Horacek,
Katherine C. Wu,
Natalia A. Trayanova,
Linwei Wang
Abstract:
Probabilistic estimation of cardiac electrophysiological model parameters serves an important step towards model personalization and uncertain quantification. The expensive computation associated with these model simulations, however, makes direct Markov Chain Monte Carlo (MCMC) sampling of the posterior probability density function (pdf) of model parameters computationally intensive. Approximated…
▽ More
Probabilistic estimation of cardiac electrophysiological model parameters serves an important step towards model personalization and uncertain quantification. The expensive computation associated with these model simulations, however, makes direct Markov Chain Monte Carlo (MCMC) sampling of the posterior probability density function (pdf) of model parameters computationally intensive. Approximated posterior pdfs resulting from replacing the simulation model with a computationally efficient surrogate, on the other hand, have seen limited accuracy. In this paper, we present a Bayesian active learning method to directly approximate the posterior pdf function of cardiac model parameters, in which we intelligently select training points to query the simulation model in order to learn the posterior pdf using a small number of samples. We integrate a generative model into Bayesian active learning to allow approximating posterior pdf of high-dimensional model parameters at the resolution of the cardiac mesh. We further introduce new acquisition functions to focus the selection of training points on better approximating the shape rather than the modes of the posterior pdf of interest. We evaluated the presented method in estimating tissue excitability in a 3D cardiac electrophysiological model in a range of synthetic and real-data experiments. We demonstrated its improved accuracy in approximating the posterior pdf compared to Bayesian active learning using regular acquisition functions, and substantially reduced computational cost in comparison to existing standard or accelerated MCMC sampling.
△ Less
Submitted 13 October, 2021;
originally announced October 2021.
-
Semi-supervised Medical Image Classification with Global Latent Mixing
Authors:
Prashnna Kumar Gyawali,
Sandesh Ghimire,
Pradeep Bajracharya,
Zhiyuan Li,
Linwei Wang
Abstract:
Computer-aided diagnosis via deep learning relies on large-scale annotated data sets, which can be costly when involving expert knowledge. Semi-supervised learning (SSL) mitigates this challenge by leveraging unlabeled data. One effective SSL approach is to regularize the local smoothness of neural functions via perturbations around single data points. In this work, we argue that regularizing the…
▽ More
Computer-aided diagnosis via deep learning relies on large-scale annotated data sets, which can be costly when involving expert knowledge. Semi-supervised learning (SSL) mitigates this challenge by leveraging unlabeled data. One effective SSL approach is to regularize the local smoothness of neural functions via perturbations around single data points. In this work, we argue that regularizing the global smoothness of neural functions by filling the void in between data points can further improve SSL. We present a novel SSL approach that trains the neural network on linear mixing of labeled and unlabeled data, at both the input and latent space in order to regularize different portions of the network. We evaluated the presented model on two distinct medical image data sets for semi-supervised classification of thoracic disease and skin lesion, demonstrating its improved performance over SSL with local perturbations and SSL with global mixing but at the input space only. Our code is available at https://github.com/Prasanna1991/LatentMixing.
△ Less
Submitted 22 May, 2020;
originally announced May 2020.
