题目：High-performance Computing, Attosecond Physics and Deep Learning

报告人：陈少豪博士（Senior Scientific Computing Consultant, Boston University）

主持人：杨小宝教授

时间：2018年6月19日（星期二）9:00

地点：物理楼（18号楼）二楼213室学术报告厅

     欢迎广大师生参加!

                  物理与光电学院

                   2018年6月12日

内容摘要：

High-performance computing (HPC) plays a key role in nowadays scientific research. Computer clusters and accelerators (such as GPU or Xeon Phi) are widely used in science and engineering. In this talk, I will first briefly introduce HPC and its recent development, then I will show applications of HPC in two research fields: (1) attosecond physics, and (2) deep learning in quantum physics.

Particularly, attosecond transient absorption (ATA) technique was used to explore ultrafast phenomenons in atomic excited states. We developed effective theoretical methods and HPC programs to deal with ATA. We analysed several interesting phenomenons in ATA spectrum, such as light-induced states, sub-cycle variation of absorption probabilities, absorption line-shape variation and its macroscopic effects. The calculation results agreed well with recent ATA experiments. Besides, we studied several interesting examples in strong-field physics, such as double ionization of atoms and filamentation of femtosecond laser pulses.

We applied a deep-learning scheme to deal with the Schrodinger Equation, which is at the heart of quantum physics. Specifically, we built a convolutional neural network model to map spatial structure of an input potential (image) to the ground-state energy of a two-dimensional helium atom. The model was learned from data. The convergence of training error and test error was obtained. The computer program is well scalable on GPUs.

报告人简介：

Shao-Hao Chen began his study in Tsinghua since 1999 and he was awarded a PhD degree in Physics in 2008 at Tsinghua University. After the defense of Chen’s PhD thesis in Jan 2008, Chen worked as a postdoc in University of Colorado and Louisiana State University from 2008 to 2014. He became permanent staff in Boston University, focusing on High Performance Computing, Parallel Computing, Computational Physics and Machine Learning and Deep Learning. Dr. Chen has published 26 peer reviewed articles in prestigious journals, with over 460 citations.