Inexact iterative numerical linear algebra for neural network-based spectral estimation and rare-event prediction

Abstract

Understanding dynamics in complex systems is challenging because there are many degrees of freedom, and those that are most important for describing events of interest are often not obvious. The leading eigenfunctions of the transition operator are useful for visualization, and they can provide an efficient basis for computing statistics such as the likelihood and average time of events (predictions). Here we develop inexact iterative linear algebra methods for computing these eigenfunctions (spectral estimation) and making predictions from a data set of short trajectories sampled at finite intervals. We demonstrate the methods on a low-dimensional model that facilitates visualization and a high-dimensional model of a biomolecular system. Implications for the prediction problem in reinforcement learning are discussed.

arXiv comment24 pages, 16 figures

References in corpus (26)

• Adam: A Method for Stochastic Optimization
• Deep learning for universal linear embeddings of nonlinear dynamics
• Identification of slow molecular order parameters for Markov model construction
• Estimating divergence functionals and the likelihood ratio by convex risk minimization
• Time-lagged autoencoders: Deep learning of slow collective variables for molecular kinetics
• Data-driven model reduction and transfer operator approximation
• Analog forecasting of extreme-causing weather patterns using deep learning
• Nonlinear Discovery of Slow Molecular Modes using State-Free Reversible VAMPnets
• Autonomous artificial intelligence discovers mechanisms of molecular self-organization in virtual experiments
• Identification of simple reaction coordinates from complex dynamics
• Galerkin Approximation of Dynamical Quantities using Trajectory Data
• Computing Committor Functions for the Study of Rare Events Using Deep Learning
• Artificial Intelligence Assists Discovery of Reaction Coordinates and Mechanisms from Molecular Dynamics Simulations
• Integrated VAC: A robust strategy for identifying eigenfunctions of dynamical operators
• Transition paths of marine debris and the stability of the garbage patches
• Predicting rare events using neural networks and short-trajectory data
• Committor Functions for Climate Phenomena at the Predictability Margin: The example of El Niño Southern Oscillation in the Jin and Timmerman model
• Coupling rare event algorithms with data-based learned committor functions using the analogue Markov chain
• Active Importance Sampling for Variational Objectives Dominated by Rare Events: Consequences for Optimization and Generalization
• Path properties of atmospheric transitions: illustration with a low-order sudden stratospheric warming model
• Computing transition path theory quantities with trajectory stratification
• Batch Stationary Distribution Estimation
• A semigroup method for high dimensional committor functions based on neural network
• Augmented Transition Path Theory for Sequences of Events
• Data-driven transition path analysis yields a statistical understanding of sudden stratospheric warming events in an idealized model
• Krylov-Bellman boosting: Super-linear policy evaluation in general state spaces

Cited by in corpus (3)

• Variational deep learning of equilibrium transition path ensembles
• Accurate estimates of dynamical statistics using memory
• Adaptive tensor train metadynamics for high-dimensional free energy exploration