Estimating Maximum Expected Value through Gaussian Approximation

46 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.

Estimating the Maximum Expected Value through Upper Confidence Bound of Likelihood (2017, background): “This assumption is following the previous work on similar analysis [7], [8].”
Estimating the Maximum Expected Value in Continuous Reinforcement Learning Problems (2017, methodology): “WE has been applied in the Weighted Q-Learning algorithm (D’Eramo, Restelli, and Nuara 2016) attaining good performance in different problems where either QLearning or Double Q-Learning performed poorly, showing superior robustness.”
Stochastic Double Deep Q-Network (2019, methodology): “In order to eliminate overestimation, various methods have been proposed, such as the double estimator operation [28], the average estimator operation [29], [30] and the weighted estimator operation [31], [32].”
Interleaved Q-Learning with Partially Coupled Training Process (2019, background): “A Gaussian estimator (GE) is proposed by [6], and the authors assume a Gaussian distribution of the sample averages, with this assumption, GE is able to reach an estimation with the bias bound smaller than that of both DE and ME.”