Skip to content
Code for the icml paper "zero inflated exponential family embedding"
Branch: master
Clone or download
Latest commit 9954ed3 Nov 2, 2017
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.gitignore add the project code Aug 6, 2017
LICENSE.txt initialize the repo Aug 6, 2017
README.md more edits of the readme file Aug 6, 2017
demo.py add the project code Aug 6, 2017
graph_builder.py correct on issue -- substract the correct vector from the vector sum … Nov 2, 2017
random_data.py add the project code Aug 6, 2017
zie.py

README.md

Zero-Inflated Exponential Family Embeddings

Introduction

This repo implements the embedding models in the 2017 ICML paper "Zero-Inflated Exponential Family Embeddings"

Zero-Inflated Exponential Family Embedding (ZIE) model is designed to learn embedding vectors of items on sparse data. It uses zero-inflated distributions as the conditional in the embedding model. Fitting a ZIE naturally downweights the zeros and dampens their influence on the model. Please see the details in the paper.

Running the code

python demo.py

Note: this repo does not contain any data -- it only use some random data to show how to use the code. The code requires numpy, scipy, and tensorflow.

Contact and cite

If you have any questions, please contact the Li-Ping Liu (liping.liulp at gmail).

If you have used the code in your work, please cite:

@inproceedings{zie17,
title = {Zero-Inflated Exponential Family Embeddings},
author = {Li-Ping Liu and David M. Blei},
booktitle ={Proceedings of the 34th International Conference on Machine Learning},
pages = {2140--2148},
year = {2017},
editor = {Doina Precup and Yee Whye Teh},
volume = {70},
series = {Proceedings of Machine Learning Research},
address = {International Convention Centre, Sydney, Australia},
month = {06--11 Aug},
publisher ={PMLR}
}

You can’t perform that action at this time.