# Classification and prediction of wave chaotic systems with machine learning techniques

## TL;DR

This paper demonstrates how supervised machine learning can classify and predict wave chaotic systems' properties from scattering data, effectively revealing hidden system details and forecasting future states.

## Contribution

It introduces machine learning methods to classify system configurations and predict future states of wave chaotic systems based on scattering data.

## Key findings

- Successfully distinguished the number of cavities in a chain from reflection data
- Developed a recurrent neural network to forecast system behavior after perturbations
- Showed machine learning can extract hidden information from complex scattering data

## Abstract

The wave properties of complex scattering systems that are large compared to the wavelength, and show chaos in the classical limit, are extremely sensitive to system details. A solution to the wave equation for a specific configuration can change substantially under small perturbations. Due to this extreme sensitivity, it is difficult to discern basic information about a complex system simply from scattering data as a function of energy or frequency, at least by eye. In this work, we employ supervised machine learning algorithms to reveal and classify hidden information about the complex scattering system presented in the data. As an example we are able to distinguish the total number of connected cavities in a linear chain of weakly coupled lossy enclosures from measured reflection data. A predictive machine learning algorithm for the future states of a perturbed complex scattering system is also trained with a recurrent neural network. Given a finite training data series, the reflection/transmission properties can be forecast by the proposed algorithm.