Frame Skip Is a Powerful Parameter for Learning to Play Atari

By Alexander Braylan, Mark Hollenbeck, Elliot Meyerson, Risto Miikkulainen. 2015

80 citations recorded by the source corpus.

Subfields: Reinforcement Learning

How researchers see this paper

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

In Reinforcement Learning

1. Sergey Levine
2. Balaraman Ravindran
3. Oriol Vinyals
4. A. Srinivas
5. Sylvain Cussat-Blanc
6. John Foley
7. Chiyuan Zhang
8. R´emi Munos

Citations in the corpus

• A Study on Overfitting in Deep Reinforcement Learning (2018, background): “Algorithms that exploit determinism of the environment by ignoring the states and memorizing effective open-loop action sequences could attain state-of-the-art scores (Braylan et al., 2015; Bellemare et al., 2015).”
• Dynamic Action Repetition for Deep Reinforcement Learning (2017, methodology): “(Braylan et al. 2015) is a work that focuses on the power of the frame skip rate hyper-parameter (it is the hyperparameter in ALE which directly controls the action repetition rate) with experiments in the Atari 2600 domain.”
• Toybox: A Suite of Environments for Experimental Evaluation of Deep Reinforcement Learning (2019, background): “Skipping frames periodically is a critically important hyperparameter for tuning algorithms to play Atari (Mnih et al., 2013; Braylan et al., 2015; Seita, 2016).”
• Evolving simple programs for playing atari games (2018, background): “An important parameter in Atari playing is frame skip [2].”