This library provides the models and the learning algorithms for learning deep stochastic stable dynamics by Invertible Flows. The model is composed by a latent stochastic stable dynamic system and an invertible flow. See 1
The models and the learning algorithms are implemented in PyTorch.
Inside the repository,
pip install -r requirements.txt
pip install -e .
Examples are placed in the examples directory.
You can run examples in
Limit Cycle with IROS dataset [1] (results are saved in examples/experiments)
python examples/train_iros.py
Goto Motions with LASA dataset [2]
python examples/train_lasa.py
Goto Motions with pouring dataset
python examples/train_pouring.py
Limit Cycle Motions with drumming dataset
python examples/train_pouring.py
Stable Dynamic Flows, named ImitationFlows in [1], represents a family of neural network architectures, which combines a latent stable dynamic system and an invertible neural network(Normalizing Flows).
You can find the set of stable dynamic system models in dynamics.
For the invertible networks, we have used RealNVP layer [3], Neural Spline Flows [4] and
Neural ODE [5]. You can find our models in flows.
[1] Julen Urain, Michele Ginesi, Davide Tateo, Jan Peters. "ImitationFlows: Learning Deep Stable Stochastic Dynamic Systems by Normalizing Flows" IEEE/RSJ International Conference on Intelligent Robots and Systems. 2020.https://arxiv.org/abs/2010.13129
[2] Khansari-Zadeh, S. Mohammad, and Aude Billard. "Learning stable nonlinear dynamical systems with gaussian mixture models." IEEE Transactions on Robotics 2011.
[3] Dinh, Laurent, Jascha Sohl-Dickstein, and Samy Bengio. "Density estimation using real nvp." International Conference on Learning Representations 2016.
[4] Durkan, C., Bekasov, A., Murray, I., & Papamakarios, G. "Neural spline flows." Advances in Neural Information Processing Systems 2019.
[5] Chen, R. T., Rubanova, Y., Bettencourt, J., & Duvenaud, D. K. "Neural ordinary differential equations". In Advances in neural information processing systems 2018.
If you found this library useful in your research, please consider citing
@article{urain2020imitationflows,
title={ImitationFlows: Learning Deep Stable Stochastic Dynamic Systems by Normalizing Flows},
author={Urain, Julen and Ginesi, Michele and Tateo, Davide and Peters, Jan},
journal={IEEE/RSJ International Conference on Intelligent Robots and Systems},
year={2020}
}
Our Flows library has been highly influenced by the amazing repositories
https://github.com/bayesiains/nsf https://github.com/rtqichen/torchdiffeq