PhMeier / Gumbel_Softmax_Actor_Critic

Gumbel Softmax in an Actor-Critic Network

Project code for the seminar 'Oldies but Goldies'.

This repository contains four files. Two files implement the basic actor-critic approach, the other two files implement the gumbelised versions.

https://github.com/DeepReinforcementLearning/DeepReinforcementLearningInAction/blob/master/Chapter%205/Ch5_book.ipynb

• First_a2c_vanilla.py: Took from: https://github.com/floodsung/a2c_cartpole_pytorch
• First_a2c_custom.py: Took from: https://github.com/floodsung/a2c_cartpole_pytorch
• Second_a2c_vanilla.py: Took from: https://github.com/DeepReinforcementLearning/DeepReinforcementLearningInAction/blob/master/Chapter%205/Ch5_book.ipynb
• Second_a2c_custom.py: Took from: https://github.com/DeepReinforcementLearning/DeepReinforcementLearningInAction/blob/master/Chapter%205/Ch5_book.ipynb