Predictive representations of state Export

@inproceedings{citeulike:3926342, abstract = {We show that states of a dynamical system can be usefully represented by multi-step, action-conditional predictions of future observations. State representations that are grounded in data in this way may be easier to learn, generalize better, and be less dependent on accurate prior models than, for example, POMDP state representations. Building on prior work by Jaeger and by Rivest and Schapire, in this paper we compare and contrast a linear specialization of the predictive approach with the state representations used in POMDPs and in k-order Markov models. Ours is the rst specic formulation of the predictive idea that includes both stochasticity and actions (controls). We show that any system has a linear predictive state representation with number of predictions no greater than the number of states in its minimal POMDP model. In predicting or controlling a sequence of observations, the concepts of state and}, author = {Littman, Michael L. and Sutton, Richard S. and Singh, Satinder}, booktitle = {In Advances In Neural Information Processing Systems 14}, citeulike-article-id = {3926342}, citeulike-linkout-0 = {http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.19.8303}, keywords = {ml}, pages = {1555--1561}, posted-at = {2009-01-23 03:24:18}, priority = {2}, title = {Predictive representations of state}, url = {http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.19.8303}, volume = {14}, year = {2002} }

TY - CONF ID - citeulike:3926342 L3 - citeulike-article-id:3926342 N2 - We show that states of a dynamical system can be usefully represented by multi-step, action-conditional predictions of future observations. State representations that are grounded in data in this way may be easier to learn, generalize better, and be less dependent on accurate prior models than, for example, POMDP state representations. Building on prior work by Jaeger and by Rivest and Schapire, in this paper we compare and contrast a linear specialization of the predictive approach with the state representations used in POMDPs and in k-order Markov models. Ours is the rst specic formulation of the predictive idea that includes both stochasticity and actions (controls). We show that any system has a linear predictive state representation with number of predictions no greater than the number of states in its minimal POMDP model. In predicting or controlling a sequence of observations, the concepts of state and EP - 1561 TI - Predictive representations of state VL - 14 T2 - In Advances In Neural Information Processing Systems 14 SP - 1555 KW - ml AU - Littman, Michael AU - Sutton, Richard AU - Singh, Satinder PY - 2002/// UR - http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.19.8303 ER -