Stable Neural Flows

Abstract

We introduce a provably stable variant of neural ordinary differential equations (neural ODEs) whose trajectories evolve on an energy functional parametrised by a neural network. Stable neural flows provide an implicit guarantee on asymptotic stability of the depth-flows, leading to robustness against input perturbations and low computational burden for the numerical solver. The learning procedure is cast as an optimal control problem, and an approximate solution is proposed based on adjoint sensivity analysis. We further introduce novel regularizers designed to ease the optimization process and speed up convergence. The proposed model class is evaluated on non-linear classification and function approximation tasks.

References in corpus (8)

• Hamiltonian Neural Networks
• Scalable Gradients for Stochastic Differential Equations
• Neural Jump Stochastic Differential Equations
• ODE${}^2$VAE: Deep generative second order ODEs with Bayesian neural networks
• Dissecting Neural ODEs
• AntisymmetricRNN: A Dynamical System View on Recurrent Neural Networks
• Graph Neural Ordinary Differential Equations
• Transport Analysis of Infinitely Deep Neural Network

Cited by in corpus (7)

• Neural Controlled Differential Equations for Irregular Time Series
• Accelerating Neural ODEs Using Model Order Reduction
• Continuous-in-Depth Neural Networks
• TorchDyn: A Neural Differential Equations Library
• Learning normalizing flows from Entropy-Kantorovich potentials
• Learning Dynamics Models with Stable Invariant Sets
• STR-GODEs: Spatial-Temporal-Ridership Graph ODEs for Metro Ridership Prediction