Learning Multimodal Transition Dynamics for Model-Based Reinforcement Learning

Code for reproducing key results in the paper Learning Multimodal Transition Dynamics for Model-Based Reinforcement Learning by Thomas M. Moerland, Joost Broekens and Catholijn M. Jonker.

Prerequisites

1. Install recent versions of:

• Python 3
• Tensorflow
• Numpy (e.g. pip install numpy)
• Matplotlib

2. Clone this repository:

git clone https://github.com/tmoer/multimodal_varinf.git

Syntax

Example:

python3 vae_main.py --logdir <logdir> --hpconfig network=1,n_rep=10,var_type='discrete',K=3,N=3,verbose=False
python3 vae_grid.py --logdir <logdir> --hpconfig network=1,n_epochs=75000,n_rep=5,var_type='continuous',z_size=8,n_flow=0,artificial_data=False,use_target_net=True,test_on_policy=True,verbose=False

For default hyper-parameters, look at the get_hps() function in the vae_grid.py and vae_main.py scripts.

Reproducing Paper Results

Run:

bash paper_toy.sh (Sec 4.1)
bash paper_grid.sh (Sec 4.2)
bash paper_grid_rl.sh (Sec 4.2)

Citation

@proceedings{moerland2017learning,
	author = "Moerland, Thomas M. and Broekens, Joost and Jonker, Catholijn M.",
	note = "arXiv preprint arXiv:1705.00470",
	journal = "Scaling Up Reinforcement Learning (SURL) Workshop @ European Machine Learning Conference (ECML)",
	title = "{Learning Multimodal Transition Dynamics for Model-Based Reinforcement Learning}",
	year = "2017"
}