Certified Feedforward Tracking for Unknown Nonlinear Systems via Invertible Neural Networks
arXiv:2608.06419
2026
Dynamics
2 ideas extracted · analyzed Aug 31, 2026
What the math gives to ML
The paper combines an invertible neural surrogate with conformal prediction to certify feedforward tracking of an unknown nonlinear dynamical system. Its transferable mechanism is to replace a difficult nonconvex inversion problem with an explicitly invertible map, then propagate a finite-sample prediction-error radius through the inverse or downstream dynamics. For neural sequence models and world models, this suggests training invertible frequency-domain surrogates and attaching conformal residual bounds that produce operational uncertainty certificates, abstention rules, or safe deployment regions rather than uncalibrated point predictions.
Ideas from this paper
✗ Failed on benchmark
2026
Attach a finite-sample conformal error radius to a neural surrogate of a dynamical or sequence model, and propagate that radius through the model's local sensitivity. The model should expose a calibrated prediction set or abstain whenever the accumulated bound exceeds a task-specific tolerance, making long-horizon failure a measurable coverage event rather than an unobserved drift.
Useful7/10
Difficulty4/10
Novelty6/10
Unverified
2026
Represent periodic input-output behavior using a compact real vector of Fourier coefficients and learn an invertible neural map from input coefficients to output coefficients. Inference then obtains the input representation for a desired periodic output by a single inverse pass instead of iterative optimization through a nonlinear forward model, while the Fourier representation reduces sequence dimensionality when high-rate signals are spectrally sparse.
Useful6/10
Difficulty6/10
Novelty4/10