Accelerated local anomaly detection via resolving attributed networks
The whole pipeline is separated into two parts:
(1) Run DSGD_CNMF.py in python, which implements CNMF and get factor matrices W and H
(2) Run DSGD_CNMF_FindOutlier.m in MATLAB to detect anomalies
(An example factorization result is in the ‘disney’ folder, in which you can evaluate directly by calling DSGD_CNMF_FindOutlier. )
The codes have been tested on Ubuntu14. Running on Windows OS is not recommended due to the instability of the multiprocessing module.
@inproceedings{liu2017accelerated,
title={Accelerated local anomaly detection via resolving attributed networks},
author={Liu, Ninghao and Huang, Xiao and Hu, Xia},
booktitle={Proceedings of the 26th International Joint Conference on Artificial Intelligence},
pages={2337--2343},
year={2017},
organization={AAAI Press}}