Modeling and Automatic Tuning in High Performance Computing
Abstract:
This literature research lays the groundwork for integrating a machine learning based algorithm selector into the molecular dynamic simulation framework AutoPas. Two refer- ence projects were consulted to analyse the performance of using machine learning models for automatic algorithm selection One is using supervised learning, the other reinforcement learning. The results though show no clear favorit among the two, as there are some key performance measurements missing for a final comparison.