MontezumaRevenge_SDRL

This is for the project "SDRL: Interpretable and Data-efficient Deep Reinforcement Learning Leveraging Symbolic Planning".

launch_training_session.sh is used for the training process, while launch_test.sh is for the test.

Requirement: Tensorflow and Keras (the experiments were run on Linux with Tensorflow version 1.3.0 and Keras version 2.1.2)

Name		Name	Last commit message	Last commit date
Latest commit History 34 Commits
trained_models_for_test		trained_models_for_test
README.md		README.md
ale_python_interface.py		ale_python_interface.py
constraint.lp		constraint.lp
environment_atari.py		environment_atari.py
first_subgoal_box.jpeg		first_subgoal_box.jpeg
goal.lp		goal.lp
hybrid_asp_dqn_training_more_dqns_new.py		hybrid_asp_dqn_training_more_dqns_new.py
hybrid_model_atari.py		hybrid_model_atari.py
hybrid_rl_il_agent_atari.py		hybrid_rl_il_agent_atari.py
hyperparameters_new.py		hyperparameters_new.py
initial.lp		initial.lp
initial_screen.jpeg		initial_screen.jpeg
launch_test.sh		launch_test.sh
launch_training_session.sh		launch_training_session.sh
libale_c.so		libale_c.so
meta_net_il.py		meta_net_il.py
montezuma_basic.lp		montezuma_basic.lp
montezuma_revenge.bin		montezuma_revenge.bin
planner.py		planner.py
q.lp		q.lp
replay_buffer.py		replay_buffer.py
result.tmp		result.tmp
schedules.py		schedules.py
segment_tree.py		segment_tree.py
show.lp		show.lp
simple_net.py		simple_net.py
tensorboard.py		tensorboard.py
test_model_more_dqns.py		test_model_more_dqns.py
visualizer.py		visualizer.py

Provide feedback