This code repository includes the source code for the Paper:

Learning to Guide Random Search
Ozan Sener, Vladlen Koltun
International Conference on Learning Representations (ICLR) 2020 

The experimentation framework is based on Ray and extends the implementation of ARS.

The source code is released under the MIT License. See the License file for details.

Please note that this is the minimal implementation of the LMRS for MuJoCo, we will update the repo with the additional code for XFoil, Pagmo, and synthetic experiments.

The code uses the following Python packages and they are required: tensorboardX, pytorch>1.0, click, numpy, torchvision, tqdm, scipy, Pillow, ray

The code is only tested in Python 3 using Anaconda environment.

If you want to run the MuJoCo experiments, install OpenAI Gym (version 0.9.3) and MuJoCo(version 0.5.7) following the instructions.

If you want to run the AirFoil experiments, install XFoil and make sure the binary is in the $PATH.

If you want to run the continous optimization benchmark, install Pagmo following esa/pagmo2.

Experiment specific parameters are provided as a json file. See the hc.json for an example.

To run an example experiment, use the command:

python mujoco_experiments.py --param_file=./hc.json

For any question, you can contact [email protected]

If you use this codebase or any part of it for a publication, please cite:

@inproceedings{ICLR2020_Sener_Koltun,
title={Learning to Guide Random Search},
author={Ozan Sener and Vladlen Koltun},
booktitle={International Conference on Learning Representations},
year={2020},
url={https://openreview.net/forum?id=B1gHokBKwS}
}