Physics-guided generative adversarial network to learn physical models
Authors:
Kazuo Yonekura
Abstract:
This short note describes the concept of guided training of deep neural networks (DNNs) to learn physically reasonable solutions. DNNs are being widely used to predict phenomena in physics and mechanics. One of the issues of DNNs is that their output does not always satisfy physical equations. One approach to consider physical equations is adding a residual of equations into the loss function; thi…
▽ More
This short note describes the concept of guided training of deep neural networks (DNNs) to learn physically reasonable solutions. DNNs are being widely used to predict phenomena in physics and mechanics. One of the issues of DNNs is that their output does not always satisfy physical equations. One approach to consider physical equations is adding a residual of equations into the loss function; this is called physics-informed neural network (PINN). One feature of PINNs is that the physical equations and corresponding residual must be implemented as part of a neural network model. In addition, the residual does not always converge to a small value. The proposed model is a physics-guided generative adversarial network (PG-GAN) that uses a GAN architecture in which physical equations are used to judge whether the neural network's output is consistent with physics. The proposed method was applied to a simple problem to assess its potential usability.
△ Less
Submitted 22 April, 2023;
originally announced April 2023.
Single-shot 3D structure determination of nanocrystals with femtosecond X-ray free electron laser pulses
Authors:
Rui Xu,
Huaidong Jiang,
Changyong Song,
Jose A. Rodriguez,
Zhifeng Huang,
Chien-Chun Chen,
Daewoong Nam,
Jaehyun Park,
Marcus Gallagher-Jones,
Sangsoo Kim,
Sunam Kim,
Akihiro Suzuki,
Yuki Takayama,
Tomotaka Oroguchi,
Yukio Takahashi,
Jiadong Fan,
Yunfei Zou,
Takaki Hatsui,
Yuichi Inubushi,
Takashi Kameshima,
Koji Yonekura,
Kensuke Tono,
Tadashi Togashi,
Takahiro Sato,
Masaki Yamamoto
, et al. (4 additional authors not shown)
Abstract:
Coherent diffraction imaging (CDI) using synchrotron radiation, X-ray free electron lasers (X-FELs), high harmonic generation, soft X-ray lasers, and optical lasers has found broad applications across several disciplines. An active research direction in CDI is to determine the structure of single particles with intense, femtosecond X-FEL pulses based on diffraction-before-destruction scheme. Howev…
▽ More
Coherent diffraction imaging (CDI) using synchrotron radiation, X-ray free electron lasers (X-FELs), high harmonic generation, soft X-ray lasers, and optical lasers has found broad applications across several disciplines. An active research direction in CDI is to determine the structure of single particles with intense, femtosecond X-FEL pulses based on diffraction-before-destruction scheme. However, single-shot 3D structure determination has not been experimentally realized yet. Here we report the first experimental demonstration of single-shot 3D structure determination of individual nanocrystals using ~10 femtosecond X-FEL pulses. Coherent diffraction patterns are collected from high-index-faceted nanocrystals, each struck by a single X-FEL pulse. Taking advantage of the symmetry of the nanocrystal, we reconstruct the 3D structure of each nanocrystal from a single-shot diffraction pattern at ~5.5 nm resolution. As symmetry exists in many nanocrystals and virus particles, this method can be applied to 3D structure studies of such particles at nanometer resolution on femtosecond time scales.
△ Less
Submitted 31 October, 2013;
originally announced October 2013.
