This repository implements the Chimera Shared Matrix Factorization (over Time) technique, first introduced in the paper "Temporally Evolving Community Detection and Prediction in Content-Centric Networks". A preprint is available at the arXiv (PDF).
This algorithm can simultaneously account for graph links, content, and temporal analysis by extracting the latent semantic structure of the network in multidimensional form, but in a way that takes into account the temporal continuity of these embeddings.
The code in this repo implements the loss function
Once optimization converges or time runs out, it will save the learning embeddings in their own files.
The code in this repository was written for Python 3 and Tensorflow.
You can install all requirements (provided you have Python 3 and are running
within a virtualenv) with pip install -r requirements.txt.
We have written a small test suite with pytest. You can run a sample prediction
with a synthetic dataset by calling py.test in the repository's root
directory.
If the code in this repository somehow helps your research, please consider citing the aforementioned paper. A BibTeX entry is provided for you below:
@inproceedings{appel2018temporally,
title={Temporally Evolving Community Detection and Prediction in COntent-Centric Networks},
author={Ana P. Appel and Renato L. F. Cunha and Charu Aggarwal and Marcela Megumi Terakado},
booktitle={Joint European Conference on Machine Learning and Knowledge Discovery in Databases},
pages={},
month = {September},
year={2018},
organization={Springer},
address = {Dublin, Ireland},
}