Bottleneck Conditional Density Estimation (ICML 2017)
Switch branches/tags
Nothing to show
Clone or download
rshu
Latest commit f5fc70f Jul 6, 2017
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.gitignore First commit Jun 30, 2017
LICENSE Initial commit Jun 29, 2017
README.md Add dependency version info Jul 7, 2017
adamax.py First commit Jun 30, 2017
config.py First commit Jun 30, 2017
data.py First commit Jun 30, 2017
extra_layers.py Remove naming of loss tensor op Jun 30, 2017
main.py Change assert to warning Jun 30, 2017
models.py Simplify sanity checks Jun 30, 2017
utils.py First commit Jun 30, 2017

README.md

bcde

The Bottleneck Conditional Density Estimator provides a semi-supervised learning framework for high-dimensional conditional density estimation. This repository provides code to run experiments in from the 2017 ICML paper Bottleneck Conditional Density Estimation.

This work was done while interning at Adobe Systems.

Dependencies

Please make sure to pip install the following dependencies

tensorflow-gpu==1.1.0
tensorbayes==0.1.1

Example

To run the model (2-layer hybrid+factored BCDE), simply do:

python main.py --model hybrid_factored