Computer Science Methods for Effective and Sustainable Simulation Studies

Abstract

Modeling and simulation (M&S) has shown to be effective at conducting what-if analyses of complex scenarios. However, the current societal and technical challenges require building increasingly complex models and carrying out larger scale simulation experiments, which call for more efficient and intelligent approaches in all aspects of simulation studies, from model creation to model execution and experimentation. In his state-of-the-art and open-challenges (STAROC) presentation on “High Performance Computing (HPC): Exploiting New Architectures”, Christopher Carothers introduced an array of new emerging hardware that supports effective and sustainable simulation studies, including ultra-dense chips, fast analog computing, wafer scale processor, low precision computing, and brain inspired chips. He went on to discuss the demand for Artificial Intelligence (AI)/Machine Learning (ML) and how it is driving the development of hardware and applications in the HPC sector. The trend of growing system complexity is also accelerating the development of software tools and hardware platforms specifically developed for AI/ML [1]. In his presentation, Carothers posed the following questions: Can AI/ML help M&S be more sustainable? Can M&S help improve AI/ML predictions?

@article{caiComputerScienceMethodsFor2023,
 author = {Wentong Cai and  Philipp Andelfinger and  Luca Bortolussi and  Christopher Carothers and  Dong Kevin Jin and  Till Köster and  Michael Lees and  Jason Liu and  Margaret Loper and  Alessandro Pellegrini and  Wen Jun Tan and Verena Wolf},
 journal = {Dagstuhl Reports},
 month = {5},
 number = {10},
 pages = {1-44},
 publisher = {Dagstuhl Reports},
 title = {Computer Science Methods for Effective and Sustainable Simulation Studies},
 volume = {12},
 year = {2023}
}