BriefGPT.xyz
Jul, 2023
基于Fenchel二次对偶的多样化离线模仿
Diverse Offline Imitation via Fenchel Duality
HTML
PDF
Marin Vlastelica, Pavel Kolev, Jin Cheng, Georg Martius
TL;DR
我们提出了一种离线技能发现算法，该算法基于互信息目标的最大化，通过KL散度进行约束，确保每种技能的状态占用率保持在离线数据集的支持范围内的专家状态占用率附近，从而学习与专家相一致的多样化技能。
Abstract
There has been significant recent progress in the area of
unsupervised skill discovery
, with various works proposing
mutual information
based objectives, as a source of intrinsic motivation. Prior works predomina
→