Data-driven predictive control of nonlinear systems using weighted regularization

arXiv:2607.09187 2026 Training 1 ideas extracted · analyzed Aug 30, 2026

What the math gives to ML

The paper replaces hard selection of historical data columns with operating-point-dependent weighted norm regularization, so nearby trajectories dominate prediction while every data column remains available. The transferable mechanism is soft localization with a strictly positive diagonal weight matrix: it preserves full-data rank and feasibility while adapting the predictor to nonlinear regime changes. A practical neural-network analogue is an online world model trained on all replay transitions, with transition losses weighted by distance from the current state-input context and a positive weight floor preventing catastrophic loss of global coverage.

Ideas from this paper

Failed on benchmark 2026

Full-Rank Local Replay

Train a neural state-space model using all replayed transitions, but assign larger weights to samples near the current operating context rather than discarding distant samples. Add a strictly positive weight floor so local adaptation cannot eliminate global coverage or make the regression problem rank-deficient. This should improve prediction across nonlinear regimes while retaining the numerical robustness of full-data training.

Useful7/10
Difficulty4/10
Novelty5/10
Paper: Data-driven predictive control of nonlinear systems using weighted regularization arXiv:2607.09187