search

notadamking / RLTrader

A cryptocurrency trading environment using deep reinforcement learning and OpenAI's gym
https://discord.gg/ZZ7BGWh
GNU General Public License v3.0
1.73k stars 539 forks source link

Ppo2 hyperparameters #126

Closed andreskull closed 5 years ago

andreskull commented 5 years ago

1) PPO2 hyperparameter ranges changed inspired by https://medium.com/aureliantactics/ppo-hyperparameters-and-ranges-6fc2d29bccbe; 2) "gamma", "ent_coef", "noptepochs": loguniform --> uniform; 3) added vf_coef and max_grad_norm