Handbook of Research on Integrating Machine Learning Into HPC-Based Simulations and Analytics

Presents recent research efforts in designing and using ML techniques on HPC systems, discusses some of the results achieved thus far by cutting-edge contributions, and highlights some of the ongoing research works in these two fields.

---

Researchers are working together to build intelligent systems that exploit a variety of real datasets using ML techniques to address problems and challenges in diverse research fields. More collaboration between the HPC and ML communities is encouraged for rapid and seamless progress toward an ecosystem that effectively serves both of these communities. As the performance improvements provided by semiconductor scaling diminish, future HPC systems are expected to exhibit an increased level of heterogeneity. These systems need to be flexible and provide low latency at all levels to effectively support new use cases and paradigms. Further, new tools and benchmarks are required to overcome the common challenges across HPC and ML applications. New programming tools, languages, compilers, and operating and runtime systems may also be needed to provide new abstractions, capabilities, and services.

This book presents to the reader recent research efforts in designing and using ML techniques on HPC systems, discusses some of the results achieved thus far by cutting-edge contributions, as well as highlights some of the ongoing research works in these two fields. Another objective is to identify research challenges and opportunities in the area spanning the intersection of HPC and ML. It is ideal for students, academics, researchers, computer scientists, computer and electrical engineers, as well as experts in the field of machine learning and high-performance computing.