The setup process for this repository requires the following steps:

Download the repository.

git clone https://github.com/raphael-group/structured-anomalies.git

• Linux/Unix
• Python (2.7 or 3.6)
• NumPy (1.17)
• SciPy (1.3)
• h5py (2.10)
• NetworkX (2.4)

• connected: This anomaly is a connected subgraph of some larger graph. To find this type of anomaly, use NetMix.
• cutsize: This anomaly is a subgraph of some larger graph such that the weight of the cut is less than some predefined rho.
• line: This anomaly is a connected subgraph of a line-shaped graph, which means it's a smaller line.
• submatrix: This anomaly is a submatrix of a square matrix.
• unconstrained: This anomaly is a subset of vertices of a graph where edges are nonexistent or ignored.

This is a .tsv file. It is needed for all anomalies except the submatrix anomaly.

Each line in this file associates a node with a score:

A    -1
B    2.5
C    3

This is a file needed only for the submatrix anomaly. It contains a square matrix of scores. An example of how to create one of these is in write_example_anomalies.py.

This is a .tsv file. It is needed only for the cutsize and connected anomalies. Each edge in this file corresponds to an edge in the network.

A    C
B    C

For most anomalies, find_anomaly.py writes to the specified output file a set of anomalous nodes corresponding to whatever structure was inputted. Each line in the output file is a node:

B
C

For the submatrix anomaly, find_anomaly.py writes a list of rows and a list of columns corresponding to the submatrix.

See the examples directory for an example of each anomaly type.