The AIverse project: Simulating, analyzing, and describing galaxies and star clusters with artificial intelligence

摘要

We present our new AIverse project in which several algorithm of artificial intelligence (AI) are used to simulate, analyze, and describe the physical properties of galaxies and star clusters. The three main purposes of the project are to (i) classify the formation and evolution processes of galaxies and star clusters, (ii) perform computer simulations in an automatic way, and (iii) convert the animation produced by the simulation into sentences using AI. Here we focus exclusively on the first component of the project as follows. We use convolutional neural networks (CNNs) to classify the formation and evolution processes of galaxies based on the two-dimensional (2D) images of galactic properties such as mass densities. This new classification method is two-stage as follows. First a large number of the synthesized 2D images of galactic properties from computer simulations are used to train a CNN for the classification. Once the CNN comes to have a very high accuracy, the CNN is then used to classify the real observational data. We discuss the effectiveness of the new classification method using the results of computer simulations on one of key formation processes of galaxies. We also discuss the number of images (Ni) required to generate from computer simulations by investigating the models with different Ni and other parameters. We briefly outline the other two components of the project and discuss their purposes.