Learning Multiple Behaviors from Unlabeled Demonstrations in a Latent Controller Space

Javier Almingol, Lui Montesano, Manuel Lopes
Proceedings of the 30th International Conference on Machine Learning, PMLR 28(3):136-144, 2013.

Abstract

In this paper we introduce a method to learn multiple behaviors in the form of motor primitives from an unlabeled dataset. One of the difficulties of this problem is that in the measurement space, behaviors can be very mixed, despite existing a latent representation where they can be easily separated. We propose a mixture model based on Dirichlet Process (DP) to simultaneously cluster the observed time-series and recover a sparse representation of the behaviors using a Laplacian prior as the base measure of the DP. We show that for linear models, e.g potential functions generated by linear combinations of a large number of features, it is possible to compute analytically the marginal of the observations and derive an efficient sampler. The method is evaluated using robot behaviors and real data from human motion and compared to other techniques.

Cite this Paper


BibTeX
@InProceedings{pmlr-v28-almingol13, title = {Learning Multiple Behaviors from Unlabeled Demonstrations in a Latent Controller Space}, author = {Almingol, Javier and Montesano, Lui and Lopes, Manuel}, booktitle = {Proceedings of the 30th International Conference on Machine Learning}, pages = {136--144}, year = {2013}, editor = {Dasgupta, Sanjoy and McAllester, David}, volume = {28}, number = {3}, series = {Proceedings of Machine Learning Research}, address = {Atlanta, Georgia, USA}, month = {17--19 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v28/almingol13.pdf}, url = {https://proceedings.mlr.press/v28/almingol13.html}, abstract = {In this paper we introduce a method to learn multiple behaviors in the form of motor primitives from an unlabeled dataset. One of the difficulties of this problem is that in the measurement space, behaviors can be very mixed, despite existing a latent representation where they can be easily separated. We propose a mixture model based on Dirichlet Process (DP) to simultaneously cluster the observed time-series and recover a sparse representation of the behaviors using a Laplacian prior as the base measure of the DP. We show that for linear models, e.g potential functions generated by linear combinations of a large number of features, it is possible to compute analytically the marginal of the observations and derive an efficient sampler. The method is evaluated using robot behaviors and real data from human motion and compared to other techniques. } }
Endnote
%0 Conference Paper %T Learning Multiple Behaviors from Unlabeled Demonstrations in a Latent Controller Space %A Javier Almingol %A Lui Montesano %A Manuel Lopes %B Proceedings of the 30th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2013 %E Sanjoy Dasgupta %E David McAllester %F pmlr-v28-almingol13 %I PMLR %P 136--144 %U https://proceedings.mlr.press/v28/almingol13.html %V 28 %N 3 %X In this paper we introduce a method to learn multiple behaviors in the form of motor primitives from an unlabeled dataset. One of the difficulties of this problem is that in the measurement space, behaviors can be very mixed, despite existing a latent representation where they can be easily separated. We propose a mixture model based on Dirichlet Process (DP) to simultaneously cluster the observed time-series and recover a sparse representation of the behaviors using a Laplacian prior as the base measure of the DP. We show that for linear models, e.g potential functions generated by linear combinations of a large number of features, it is possible to compute analytically the marginal of the observations and derive an efficient sampler. The method is evaluated using robot behaviors and real data from human motion and compared to other techniques.
RIS
TY - CPAPER TI - Learning Multiple Behaviors from Unlabeled Demonstrations in a Latent Controller Space AU - Javier Almingol AU - Lui Montesano AU - Manuel Lopes BT - Proceedings of the 30th International Conference on Machine Learning DA - 2013/05/26 ED - Sanjoy Dasgupta ED - David McAllester ID - pmlr-v28-almingol13 PB - PMLR DP - Proceedings of Machine Learning Research VL - 28 IS - 3 SP - 136 EP - 144 L1 - http://proceedings.mlr.press/v28/almingol13.pdf UR - https://proceedings.mlr.press/v28/almingol13.html AB - In this paper we introduce a method to learn multiple behaviors in the form of motor primitives from an unlabeled dataset. One of the difficulties of this problem is that in the measurement space, behaviors can be very mixed, despite existing a latent representation where they can be easily separated. We propose a mixture model based on Dirichlet Process (DP) to simultaneously cluster the observed time-series and recover a sparse representation of the behaviors using a Laplacian prior as the base measure of the DP. We show that for linear models, e.g potential functions generated by linear combinations of a large number of features, it is possible to compute analytically the marginal of the observations and derive an efficient sampler. The method is evaluated using robot behaviors and real data from human motion and compared to other techniques. ER -
APA
Almingol, J., Montesano, L. & Lopes, M.. (2013). Learning Multiple Behaviors from Unlabeled Demonstrations in a Latent Controller Space. Proceedings of the 30th International Conference on Machine Learning, in Proceedings of Machine Learning Research 28(3):136-144 Available from https://proceedings.mlr.press/v28/almingol13.html.

Related Material