Skip to content

shihgianlee/openai-lunar-lander

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Lunar Lander

Lunar Lander

This is an attempt to solve OpenAI Lunar Lander-v2 using Deep Reinforcement Learning.

Implementation

The search for hyperparameters values are challenging because of the large hyperparameters space need to be searched. As a result, we use the hyperparameters values from Deep Q-Learning with Keras and Gym that is used to solve Cartpole-v1 as starting point. The lunar_lander.py file has the training code for lunar lander model. For the longest time, the rewards were hovering between 0 and negative territories. The breakthrough came when we replace epsilon-greedy exploration strategy with Boltzman exploration strategy.

Result

Lunar Lander rewards

Credits

Credits are given in the source and References.

References

[1] Deep Q-Learning with Keras and Gym. URL: https://keon.io/deep-q-learning/

[2] Playing Atari with Deep Reinforcement Learning. URL: https://www.cs.toronto.edu/~vmnih/docs/dqn.pdf

[3] Artificial Intelligence: Representation and Problem. URL: https://www.cs.cmu.edu/afs/cs/academic/class/15381-s07/www/slides/050107reinforcementLearning1.pdf

[4] Human-level control through deep reinforcement learning. URL: https://web.stanford.edu/class/psych209/Readings/MnihEtAlHassibis15NatureControlDeepRL.pdf

[5] Reinforcement Learning w/ Keras + OpenAI: DQNs. URL: https://towardsdatascience.com/reinforcement-learning-w-keras-openai-dqns-1eed3a5338c

[6] Keras RL. URL: https://github.com/keras-rl/keras-rl

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages