MIMI is an algorithm for training an interface to map user command signals to system actions through unsupervised human-in-the-loop reinforcement learning.
mimi
into your home directory ~
mimi/
conda create -n mimienv python=3.6
pip install -r requirements.txt
and pip install pyglet==1.5.11
your_install_dir/gym/envs/box2d/lunar_lander.py
with deps/box2d/lunar_lander.py
mimi
package with python setup.py install
mimi/notebooks
provide an entry-point to the code base, where you can
play around with the environments and reproduce the figures from the paper.If you find this software useful in your work, we kindly request that you cite the following paper:
@article{mimi2022,
title={First Contact: Unsupervised Human-Machine Co-Adaptation via Mutual Information Maximization},
author={Reddy, Siddharth and Levine, Sergey and Dragan, Anca D.},
journal={arXiv preprint arXiv:2205.12381},
year={2022}
}
Explore the latent space of a generative model of MNIST images using hand gestures