跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2611.21465 · v1

Warm Starting Neural Dynamics Models From Few Jacobian Samples

用少量雅可比样本热启动神经动力学模型

首次提交 2026-11-17 01:46
版本更新 2026-11-17 01:46
雷达观测 2026-11-17 16:10
arXiv ↗PDF ↗

摘要

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.

AI 论文解读

基于摘要 · 非全文审阅

提出 TDec-Init,利用少量转移雅可比样本通过张量分解解耦出权重与非线性,从而热启动神经网络动力学模型。

正在读取账号与研究记录…

版本差异

目前只收录了一个版本。收录下一版后,可以对照标题、作者、分类和摘要。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v1 · 2026-11-17 01:46 · 最新观测版本

Warm Starting Neural Dynamics Models From Few Jacobian Samples

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.

查看此版本解读 →查看此版本原文 ↗