Learning Predictive State Representations

By Satinder Singh, Michael L. Littman, Nicholas K. Jong, D. Pardoe, Peter Stone. 2003

148 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Deep spatial autoencoders for visuomotor learning (2015, background): “Prior work has proposed to learn representations using general priors from the physical world [2] or the ability to predict future observations [18], [19], [20].”
Manifold Embeddings for Model-Based Reinforcement Learning under Partial Observability (2009, methodology): “The question of state space representation in partially observable domains was tackled under the POMDP framework [14] and recently in the PSR framework [19].”
Bootstrap Latent-Predictive Representations for Multitask Reinforcement Learning (2020, background): “Littman et al. (2001); Singh et al. (2003) introduced predictive state representations (PSRs) and an algorithm for learning such representations.”
Predictive State Representations: A New Theory for Modeling Dynamical Systems (2004, background): “Recent work on PSRs has begun to theoretically and empirically explore these advantages (Singh, Littman, Jong, Pardoe, & Stone, 2003; James & Singh, 2004).”
Learning predictive state representations using non-blind policies (2006, methodology): “A number of algorithms have been successfully shown to perform both discovery and learning (Singh et al., 2003; James & Singh, 2004; Rosencrantz et al., 2004; Wolfe et al., 2005; Wiewiora, 2005; McCracken & Bowling, 2006).”
Blind Construction of Optimal Nonlinear Recursive Predictors for Discrete Sequences (2004, background): “Causal states are also closely related to the “predictive state representations” (PSRs) of controlled dynamical systems due to Littman, Sutton and Singh [9], though PSRs are not generally minimal.”
Probabilistic Recurrent State-Space Models (2018, background): “Representations range from SSMs (Van Overschee & De Moor, 2012) over Predictive State Representations (PSRs) (Littman & Sutton, 2002; Singh et al., 2003; Rudary & Singh, 2004) to autoregressive models (Murray-Smith & Girard, 2001; Girard et al.”
Universal Value Function Approximators (2015, methodology): “Second, generalized value functions can be used as features to represent state (Schaul & Ring, 2013); this is a form of predictive representation (Singh et al., 2003).”
Learning Visual Feature Spaces for Robotic Manipulation with Deep Spatial Autoencoders (2015, background): “Prior work has proposed to learn representations using general priors from the physical world [2] or the ability to predict future observations [18], [19], [20].”
Exponential Family Predictive Representations of State (2007, background): “Singh et al.(2003) presented the first algorithm for the learning algorithm, which assumes thatthe core tests are given and uses a gradient algorithm to solve the learning problem.”