LEAD PI (S): DR. HAOHUAN FU
The goal is to build the bridges between the physics-based numerical models and the data-oriented models learned through training processes, so as to enable the interaction and the mutual enhancement between the two. Based on existing work, student projects will continue the effort based on the world-leading supercomputers, and investigate:
- tools and methods to support complex coupled simulations that cover multiple scales and multiple processes required to capture the key scientific factors, and
- big data analysis platform that can intelligently process huge volumes of scientific data.
Participants would then investigate new methods that can apply learning networks trained from observation data sets to calibrate, iteratively improve, and to enhance complex numerical models. By using such methods, we can add new types of data, such as globally-distributed seismic station records, climate reanalysis data, high resolution remote sensing data, and historical literature, into earthquake and climate modeling software.
Students applying to this project are expected to have:
- Required: Experience programming in Fortran, C/C++ or a comparable language
- Required: Experience using clusters and/or HPC systems
- Required: Strong background in computer architecture and numerical computing
- Hardware: The SunwayTaihuLight Supercomputer using Sunway Raise OS 2.0.5 based on Linux as the operating system
- Software Stack: includes basic compiler components, such as C/C++, and Fortran compilers, an automatic vectorization tool, and basic math libraries. The Sunway OpenACC, a customized parallel compilation tool that supports OpenACC 2.0 syntax and targets the SW26010 many-core processor, is also available.