Original title: Alpha Zero for Physics: Application of Symbolic Regression with Alpha Zero to find the analytical methods in physics
Authors: Yoshihiro Michishita
In the world of machine learning, neural networks have become superstars, solving everything from language to games. But in physics, finding analytical methods through machine learning hasn’t been explored much. That’s where Alpha Zero steps in. This paper introduces a new idea called Alpha Zero for physics (AZfP). They propose using symbolic regression with Alpha Zero to craft analytical methods for physics problems. It’s like teaching a computer to discover complex equations from scratch. To prove its power, they showcase how AZfP nails down high-frequency expansion in Floquet systems. This breakthrough hints at AZfP’s potential to create a whole new theoretical framework in physics. It’s like giving machines the ability to crack open the secrets of the universe by learning the language of its laws.
Original article: https://arxiv.org/abs/2311.12713