Poster Presentation

Interpreting N-body simulation using machine-learning methods

Mario Pasquato (Yonsei University, Korea)

Recently, numerical simulations (especially N-body and Fokker-Planck) have come to dominate the field of Globular Cluster (GC) dynamics. With these techniques and the latest hardware it is relatively easy to simulate the dynamical evolution of a cluster over its lifetime in a reasonable time frame. However, while simulations are constantly improving, the understanding of the data they produce is still problematic. To simulate is not necessarily to understand, and extracting information from the output of large simulations is a difficult business, that usually comes down to the individual ability of the scientist interpreting the results. In order to reduce this asymmetry between the production of data from simulations and their interpretation, I explore several different methods to automatically interpret simulation data, all based on machine learning techniques. I discuss each technique and its application to GC simulations through a concrete example where I address a scientific question regarding GC evolution.