arXiv · 2611.21465
Warm Starting Neural Dynamics Models From Few Jacobian Samples
Abstract
Model-based reinforcement learning (MBRL) leverages a model of the environment, used for planning and generating samples without interacting with the real world. Motivated by evidence that informed initialization significantly improves the efficiency of MBRL, we propose TDec-Init, an algorithm to warm-start neural dynamics models using transition Jacobians. Using a decoupling algorithm based on the decomposition of a tensor of stacked Jacobian samples, we recover weights and flexible nonlinearities, and use them to initialize a feedforward neural network. In the low-data regime, we show on several benchmark environments that our approach outperforms random initialization and value training, and either matches or outperforms Sobolev training, while requiring fewer training steps between episodes.
Explore related subjects
Keep this discovery
Explore connections, maps & timelines
Maxime Rochkoulets, Joppe De Jonghe, Mariya Ishteva. 2026-11-16. Warm Starting Neural Dynamics Models From Few Jacobian Samples. https://arxiv.science/abs/2611.21465
Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite arXiv Science by Indiciae: BibTeX · RIS.