This is our project for CSE 571. We implemented Deep Ensemble, Differentiable Particle Filter and a simple behavior cloning algorithm on cartpole system.

• The master branch is for DPF. For the other two algorithms, please checkout to the corresponding branch.
• To generate data for training, please checkout the data/ folder for more details.

Adapted Tensorflow code from this repo.
Tested on sinusoidal toy dataset.

Original implementation from the authors.

We provided pretrained models for DPF, to run the recurrent model

python cartpole_test.py --mode rec

run the end-to-end model

python cartpole_test.py --mode e2e