# Geometry and Determinism of Optimal Stationary Control in Partially Observable Markov Decision Processes

@article{Montfar2015GeometryAD, title={Geometry and Determinism of Optimal Stationary Control in Partially Observable Markov Decision Processes}, author={Guido Mont{\'u}far and Keyan Zahedi and Nihat Ay}, journal={ArXiv}, year={2015}, volume={abs/1503.07206} }

It is well known that for any finite state Markov decision process (MDP) there is a memoryless deterministic policy that maximizes the expected reward. For partially observable Markov decision processes (POMDPs), optimal memoryless policies are generally stochastic. We study the expected reward optimization problem over the set of memoryless stochastic policies. We formulate this as a constrained linear optimization problem and develop a corresponding geometric framework. We show that any POMDP… CONTINUE READING

4

Twitter Mentions

#### Citations

##### Publications citing this paper.

SHOWING 1-4 OF 4 CITATIONS

## Geometry of Policy Improvement

VIEW 7 EXCERPTS

CITES BACKGROUND

## Open Problem: Approximate Planning of POMDPs in the class of Memoryless Policies

VIEW 3 EXCERPTS

CITES BACKGROUND

## Trust Region Policy Optimization of POMDPs

VIEW 3 EXCERPTS

CITES BACKGROUND

#### References

##### Publications referenced by this paper.

SHOWING 1-10 OF 24 REFERENCES

## Infinite-Horizon Policy-Gradient Estimation

VIEW 2 EXCERPTS

## Information-theoretic approach to interactive learning

VIEW 1 EXCERPT

## Reinforcement Learning: An Introduction

VIEW 1 EXCERPT