GitHub - dcbiton/olfactory-search-rl-lstm: This repository shows one sample implementation of using a Recurrent Neural Network (RNN) architecture, specifically the Long Short Term Memory RNN (LSTM) in Reinforcement Learning (RL). The LSTM is used in RL to demonstrate one way of employing memory to the learning agent in order to efficiently remember the important signals from the environment and successfully complete the task while maximizing the reward. The main source of this notebook is the paper by Bram Bakker in 2002 entitled Reinforcement Learning with Long Short-Term memory. However, the main focus is on the incorporation of LSTM in Q-learning in the T-maze task. Some important details will be explained as much as possible but not everything will be revealed

dcbiton / olfactory-search-rl-lstm Public

Notifications You must be signed in to change notification settings
Fork 0
Star 2

Notifications

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
README.md		README.md
RL_LSTM_class_updown.py		RL_LSTM_class_updown.py
RL_LSTM_env_updown.py		RL_LSTM_env_updown.py
evaluate_model_fcn_updown.py		evaluate_model_fcn_updown.py
main.py		main.py
main_fcn_updown.py		main_fcn_updown.py
pytorch_p37_env.yaml		pytorch_p37_env.yaml

Repository files navigation

This repository shows one sample implementation of using a Recurrent Neural Network (RNN) architecture, specifically the Long Short Term Memory RNN (LSTM) in Reinforcement Learning (RL). The LSTM is used in RL to demonstrate one way of employing memory to the learning agent in order to efficiently remember the important signals from the environment and successfully complete the task while maximizing the reward. The main source of this notebook is the paper by Bram Bakker in 2002 entitled Reinforcement Learning with Long Short-Term memory. However, the main focus is on the incorporation of LSTM in Q-learning in the T-maze task. Some important details will be explained as much as possible but not everything will be revealed