Performance bounds for λ policy iteration and application to the game of Tetris

23 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.

Approximate Policy Iteration Schemes: A Comparison (2014, background): “As a matter of fact, several other dynamic programming algorithms, namely AVI (Munos, 2007), λ PI (Scherrer, 2013), AMPI (Scherrer et al., 2012), come with guarantees involv-7 ing the worst constant C (2 , 1 , 0) , which suggests that they should not be competitive with the best algorithms we have…”
Beyond the One Step Greedy Approach in Reinforcement Learning (2018, background): “This generalization is similar to the λ -PI which interpolates between standard PI and VI (Bertsekas & Ioffe, 1996; Bertsekas & Tsitsiklis, 1995; Scherrer, 2013b; Bertsekas, 2015).”
Approximate Dynamic Programming Finally Performs Well in the Game of Tetris (2013, unknown): “More recently, Thiery and Scherrer [20] selected a set of 9 features (including those of Dellacherie’s) and optimized the weights with the CE method.”
A Lyapunov-based Approach to Safe Reinforcement Learning (2018, background): “While in the worst case SVI converges in K = O(T) steps [7] and SPI converges in K = O(|X ||A|T logT) steps [37], in practice, K is much smaller than |X ||A|.”