Skip to content
No description, website, or topics provided.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
hodge_experiment
synthetic_data
vector_field_deformations
.gitignore
README.md

README.md

conservative_autoencoder

Python (Theano) implementation of conservative auto-encoder code provided by Daniel Jiwoong Im and Mohamed Ishmael Diwan Belghazi. The codes include experiments on hodge decomposition, in particular convservative components (for now), and vector field deformations in 2D. For more information, see

@article{Im2015cae,
    title={Conservativeness of untied auto-encoders},
    author={Im, Daniel Jiwoong and Belghanzi, Mohamed Ishmael Diwan and Memisevic, Roland},
    journal={AAAI-16 AAAI Conference on Artificial Intelligence},
    year={2015}
}

If you use this in your research, we kindly ask that you cite the above workshop paper

Dependencies

Packages

How to run

Entry code for one-bit flip and factored minimum probability flow for mnist data are

    - /hodge_experiment/mnist_synthetic.py
    - /vector_field_deformation/mnist_train.py
You can’t perform that action at this time.