On discovery and learning of models with predictive representations of state for agents with continuous actions and observations
By David Wingate, Satinder Singh. 2007
23 citations recorded by the source corpus.
Subfields: Artificial Intelligence, Control and Systems Engineering
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Sergey Levine
- Chelsea Finn
- Pieter Abbeel
- Byron Boots
- Dieter Fox
- Trevor Darrell
- Ruslan Salakhutdinov
- Yoshua Bengio
In Control and Systems Engineering
- Sergey Levine
- Byron Boots
- Dieter Fox
- Geoffrey J. Gordon
- Arunkumar Byravan
- Chelsea Finn
- Sajid M. Siddiqi
- Joschka Boedecker
Citations in the corpus
- Hilbert Space Embeddings of Predictive State Representations (2013, neutral): “Previous approaches attempt to learn continuous PSRs by leveraging kernel density estimation [4] or modeling PSR distributions with exponential families [5, 6]; each of these methods must contend with drawbacks such as slow rates of statistical convergence and difficult numerical integration.”
- Closing the learning-planning loop with predictive state representations (2009, methodology): “continuous observations in PSRs is not new [29], extending our algorithm to use KDE results, for the first time, in a statistically consistent learning algorithm for PSRs with continuous observations.”
- Exponential Family Predictive Representations of State (2007, background): “Recent work has introduced the more general not ion f using parameters that model the distribution of lengthn futures as the statistics of interest [8, 12].”
- Construction of approximation spaces for reinforcement learning (2013, background): “Extensions to continuous state and observation spaces are rare (Wingate and Singh, 2007) and linear approximation not straight forward.”
- Autonomous Learning of State Representations for Control: An Emerging Field Aims to Autonomously Learn State Representations for Reinforcement Learning Agents from Their Real-World Sensor Observations (2015, background): “There are several possibilities to make Z Markov, for example temporal embedding [37], liquid state machines [31], belief states [18] or predictive state representations [29, 55], which each come with their own disadvantages.”
- Learning predictive models of a depth camera & manipulator from raw execution traces (2014, background): “Unfortunately, evaluation of PSRs have long been restricted to learning fairly simple “grid-world” type simulated environments [10], [12]–[14], a fact which has lead to the perception that PSRs are not flexible enough, or that PSR learning algorithms are not efficient enough, to represent high-dimen”