AutoML for Fast Simulation

The extensive program of high energy physics (HEP) experiments relies on Monte Carlo (MC) simulation. MC consists of a basis for testing hypothesis of the underlying distribution of the data. The need for fast and large scale simulated samples for HEP experiments motivates the development of new simulation techniques. Some of those techniques, belonging to a so-called Fast Simulation, are based on Machine Learning (ML) algorithms. They use a set of training variables known as "hyperparameters". The choice of these parameters can significantly affect the model's performance. Automated machine learning (AutoML) allows to lift the burden of a manual optimization. The goal of this project is to understand how to use AutoML for fast simulation, specifically for shower simulation in the calorimeter. We implemented AutoML to tune the hyperparameters of a ML model for shower simulation. After, a metric was defined to evaluate the model's performance, combining physics and ML validations. This metric was then used to test and compare the existing AutoML techniques