Learning particle interaction rules from motion data

Paper Title:

Data-Driven Model Selections of Second-Order Particle Dynamics via Integrating Gaussian Processes with Low-Dimensional Interacting Structures

Published on:

1 November 2023

Primary Category:

Machine Learning

Paper Authors:

Jinchao Feng,

Charles Kulick,

Sui Tang

Bullets

Key Details

•

A Gaussian process framework learns particle interaction models and governing equations from data

•

It identifies interaction types and system order with only scarce, noisy trajectory data

•

Efficient techniques are developed to improve computational scalability

•

The method is demonstrated on modeling real fish schooling data

Explore the topics in this paper

fish schooling

gaussian processes

interaction models

particle systems

trajectory data

AI generated summary

Learning particle interaction rules from motion data

This paper develops a Gaussian process-based approach to learn interaction models for particle systems from scarce, noisy trajectory data. It can identify if interactions depend on positions, velocities, or both, and if the system is first or second order. The method is applied to model real fish schooling data.