Timezone: »

Refactoring Policy for Compositional Generalizability using Self-Supervised Object Proposals
Tongzhou Mu · Jiayuan Gu · Zhiwei Jia · Hao Tang · Hao Su

Mon Dec 07 09:00 PM -- 11:00 PM (PST) @ Poster Session 0 #148

We study how to learn a policy with compositional generalizability. We propose a two-stage framework, which refactorizes a high-reward teacher policy into a generalizable student policy with strong inductive bias. Particularly, we implement an object-centric GNN-based student policy, whose input objects are learned from images through self-supervised learning. Empirically, we evaluate our approach on four difficult tasks that require compositional generalizability, and achieve superior performance compared to baselines.

Author Information

Tongzhou Mu (University of California, San Diego)
Jiayuan Gu (University of California, San Diego)
Zhiwei Jia (University of California, San Diego)
Hao Tang (Shanghai Jiao Tong University)
Hao Su (UCSD)

More from the Same Authors