TorchLean API

NN.MLTheory.LearningTheory.Stability.Dynamics

Stability of dynamical systems #

This entrypoint collects the discrete-time stability layer for maps of the form x_{t+1} = f x_t and input-driven systems x_{t+1} = f x_t u_t.

The spec file states the mathematical predicates: Lyapunov stability, asymptotic stability, exponential stability, input-to-state stability, BIBO stability, incremental stability, practical stability, finite-time stability, and data/model stability. The runtime file provides Float diagnostics for concrete systems. As with robustness, the diagnostic layer is empirical unless a separate theorem connects it to a certified bound.

References: