https://github.com/abojchevski/paican

Implementation of "Bayesian Robust Attributed Graph Clustering: Joint Learning of Partial Anomalies and Group Structure".
https://github.com/abojchevski/paican

anomalies bayesian clustering community-detection graphs robust tensorflow

Last synced: about 2 months ago
JSON representation

Implementation of "Bayesian Robust Attributed Graph Clustering: Joint Learning of Partial Anomalies and Group Structure".

• Host: GitHub
• URL: https://github.com/abojchevski/paican
• Owner: abojchevski
• License: mit
• Created: 2018-04-04T14:38:05.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2020-05-05T19:18:25.000Z (over 4 years ago)
• Last Synced: 2024-06-10T04:38:02.134Z (3 months ago)
• Topics: anomalies, bayesian, clustering, community-detection, graphs, robust, tensorflow
• Language: Python
• Homepage: https://www.kdd.in.tum.de/paican
• Size: 6.99 MB
• Stars: 24
• Watchers: 2
• Forks: 13
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

• awesome-community-detection - [Python Reference

README

        
# Paican



Tensorflow implementation of the method proposed in the paper: "Bayesian Robust Attributed Graph Clustering: Joint Learning of Partial Anomalies and Group Structure", Aleksandar Bojchevski and Stephan Günnemann, AAAI 2018.

## Installation

```bash

python setup.py install

```

## Requirements

* tensorflow (>=1.4, <=2.0)

* sklearn (only for evaluation)

Note: If you are using tensorflow >=2.0 you can stull run the above code by replacing the tensorflow import with

```

import tensorflow.compat.v1 as tf

tf.disable_v2_behavior()

```

and changing `tf.contrib.distributions` to `tf.distributions`. 

## Data

Each of the dataset folders consists of the following files:

* A.mtx - the adjacency matrix in scipy's sparse csr_matrix format

* X.mtx - the attribute matrix in scipy's sparse csr_matrix format

* feature_to_index.npy - a dictionary mapping a feature label to index (e.g. 'neurology' -> 5)

* node_to_index.npy - a dictionary mapping a node label (e.g name of a person or paper ID) to index

* z.npy - ground truth clusters if available

* label_to_cluster.npy - a dictionary mapping a label (e.g. journal, party) to cluster index

## Demo

* See the notebook example.ipynb for a simple demo.

* Visit [our website](https://www.kdd.in.tum.de/paican) for an interactive plot that shows the inferred clustering on a subset of the Amazon dataset.

## Cite

Please cite our paper if you use this code in your own work.