Plan2Vec: Unsupervised Representation Learning by Latent Plans

Ge Yang, Amy Zhang, Ari Morcos, Joelle Pineau, Pieter Abbeel, Roberto Calandra
Proceedings of the 2nd Conference on Learning for Dynamics and Control, PMLR 120:935-946, 2020.

Abstract

In this paper, we introducePlan2Vec, an model-based method to learn state representation fromsequences of off-policy observation data via planning. In contrast to prior methods, plan2vec doesnot require grounding via expert trajectories or actions, opening it up to many unsupervised learningscenarios. When applied to control, plan2vec learns a representation that amortizes the planningcost, enabling test time planning complexity that is linear in planning depth rather than exhaustiveover the entire state space. We demonstrate the effectiveness of Plan2Vec on one simulated andtwo real-world image datasets, showing that Plan2Vec can effectively acquire representations thatcarry long-range structure to accelerate planning. Additional results and videos can be found athttps://sites.google.com/view/plan2vec

Cite this Paper


BibTeX
@InProceedings{pmlr-v120-yang20b, title = {Plan2Vec: Unsupervised Representation Learning by Latent Plans}, author = {Yang, Ge and Zhang, Amy and Morcos, Ari and Pineau, Joelle and Abbeel, Pieter and Calandra, Roberto}, booktitle = {Proceedings of the 2nd Conference on Learning for Dynamics and Control}, pages = {935--946}, year = {2020}, editor = {Bayen, Alexandre M. and Jadbabaie, Ali and Pappas, George and Parrilo, Pablo A. and Recht, Benjamin and Tomlin, Claire and Zeilinger, Melanie}, volume = {120}, series = {Proceedings of Machine Learning Research}, month = {10--11 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v120/yang20b/yang20b.pdf}, url = {https://proceedings.mlr.press/v120/yang20b.html}, abstract = {In this paper, we introducePlan2Vec, an model-based method to learn state representation fromsequences of off-policy observation data via planning. In contrast to prior methods, plan2vec doesnot require grounding via expert trajectories or actions, opening it up to many unsupervised learningscenarios. When applied to control, plan2vec learns a representation that amortizes the planningcost, enabling test time planning complexity that is linear in planning depth rather than exhaustiveover the entire state space. We demonstrate the effectiveness of Plan2Vec on one simulated andtwo real-world image datasets, showing that Plan2Vec can effectively acquire representations thatcarry long-range structure to accelerate planning. Additional results and videos can be found athttps://sites.google.com/view/plan2vec} }
Endnote
%0 Conference Paper %T Plan2Vec: Unsupervised Representation Learning by Latent Plans %A Ge Yang %A Amy Zhang %A Ari Morcos %A Joelle Pineau %A Pieter Abbeel %A Roberto Calandra %B Proceedings of the 2nd Conference on Learning for Dynamics and Control %C Proceedings of Machine Learning Research %D 2020 %E Alexandre M. Bayen %E Ali Jadbabaie %E George Pappas %E Pablo A. Parrilo %E Benjamin Recht %E Claire Tomlin %E Melanie Zeilinger %F pmlr-v120-yang20b %I PMLR %P 935--946 %U https://proceedings.mlr.press/v120/yang20b.html %V 120 %X In this paper, we introducePlan2Vec, an model-based method to learn state representation fromsequences of off-policy observation data via planning. In contrast to prior methods, plan2vec doesnot require grounding via expert trajectories or actions, opening it up to many unsupervised learningscenarios. When applied to control, plan2vec learns a representation that amortizes the planningcost, enabling test time planning complexity that is linear in planning depth rather than exhaustiveover the entire state space. We demonstrate the effectiveness of Plan2Vec on one simulated andtwo real-world image datasets, showing that Plan2Vec can effectively acquire representations thatcarry long-range structure to accelerate planning. Additional results and videos can be found athttps://sites.google.com/view/plan2vec
APA
Yang, G., Zhang, A., Morcos, A., Pineau, J., Abbeel, P. & Calandra, R.. (2020). Plan2Vec: Unsupervised Representation Learning by Latent Plans. Proceedings of the 2nd Conference on Learning for Dynamics and Control, in Proceedings of Machine Learning Research 120:935-946 Available from https://proceedings.mlr.press/v120/yang20b.html.

Related Material