Deep learning in bifurcations of particle trajectories

Abstract

We show that deep learning algorithms can be deployed to study bifurcations of particle trajectories. We demonstrate this for two physical systems, the unperturbed Duffing equation and charged particles in magnetic reversal by using the AI Poincaré algorithm. We solve the equations of motion by using a fourth-order Runge-Kutta method to generate a dataset for each system. We use a deep neural network to train the data. A noise characterized by a noise scale L is added to data during the training. By using a principal component analysis, we compute the explained variance ratios for these systems which depend on the noise scale. By plotting explained ratios against the noise scale, we show that they change at bifurcations. For different values of the Duffing equation parameters, these changes are of the form of different patterns of growth-decline of the explained ratios. For the magnetic reversal, the changes are of the form of a change in the number of principal components. We comment on the use of this technique for other dynamical systems with bifurcations.

References in corpus (15)

• Machine learning and the physical sciences
• A high-bias, low-variance introduction to Machine Learning for physicists
• Principal Component Analysis: A Natural Approach to Data Exploration
• Machine Learning in Nuclear Physics
• AI Poincaré: Machine Learning Conservation Laws from Trajectories
• AI Poincaré 2.0: Machine Learning Conservation Laws from Differential Equations
• A one-dimensional Vlasov-Maxwell equilibrium for the force-free Harris sheet
• Contemporary machine learning: a guide for practitioners in the physical sciences
• Discovering Symmetry Invariants and Conserved Quantities by Interpreting Siamese Neural Networks
• Fundamental problems in statistical physics XIV: Lecture on Machine Learning
• Neural Empirical Bayes
• Machine learning prediction for mean motion resonance behaviour -- The planar case
• Chaos in the quantum Duffing oscillator in the semiclassical regime under parametrized dissipation
• Regimes of charged particle dynamics in current sheets: the machine learning approach
• Discovering New Interpretable Conservation Laws as Sparse Invariants