Visualizing Overlapping Biclusterings and Boolean Matrix Factorizations

README updated on the 27/06/2023

This archive contains the code to visualize biclusters, as described in the paper:

Thibault Marette, Pauli Miettinen, and Stefan Neumann (2023). Visualizing Overlapping Biclusterings and Boolean Matrix Factorizations. In Machine Learning and Knowledge Discovery in Databases: Research Track - European Conference, ECML PKDD 2023, Turin, Italy, September 18-22, 2023, Proceedings, Part I (pp. 743–758). Springer.

You are free to use and edit the code, as long as you cite us using the bibtex snippet below:

@inproceedings{marette2023visualizing,
  author       = {Thibault Marette and
                  Pauli Miettinen and
                  Stefan Neumann},
  title        = {Visualizing Overlapping Biclusterings and Boolean Matrix Factorizations},
  booktitle    = {Machine Learning and Knowledge Discovery in Databases: Research Track
                  - European Conference, {ECML} {PKDD} 2023, Turin, Italy, September
                  18-22, 2023, Proceedings, Part {I}},
  series       = {Lecture Notes in Computer Science},
  volume       = {14169},
  pages        = {743--758},
  publisher    = {Springer},
  year         = {2023},
  doi          = {10.1007/978-3-031-43412-9\_44},
}

If you want to use basso clustering algorithm, extract includes/basso-0.5.tar.gz. Once the archive is extracted, follow the installation instructions at includes/basso-0.5/README.

For Python packages requirements, see requirements.txt.

Execute the code

The code is available in the folder biclusteringVisualization/ Code examples are available in demo.py, demoAdvanced.py and demoAdvanced2.py

Documentation

visualizeClustering.visualizeClustering(
    inputFile,
    outputFile,
    clusteringAlgorithm="PCV",
    nbClusters=5,
    rowClusters="",
    columnClusters="",
    inputIsList=False,
    orderingMethod="TSPheuristic", 
    heuristicTime=20,
    plotUnorderedMatrix=False,
    showObjectiveFunctionValues=False
    )

Parameters:

• inputFile: string

  Location of the input file

• outputFile: string

  Location of the output file

• clusteringAlgorithm: string

  Biclustering algorithm to use. PCV (default) and basso are supported.

• nbClusters: int

  Number of clusters for the clustering algorithm. Default value is 5.

• rowClusters: string

  location of the file for the clustering of the rows. Optionnal (only use if you do not want to use clustering algorithm)

• columnClusters: string | list

  location of the file for the clustering of the columns. Optionnal (only use if you do not want to use clustering algorithm)

• orderingMethod: string | list

  Algorithm to use for the ordering of the matrix in the visualization. Default value is TSPHeuristic. ADVISER, as well as the greedy algorithms used for the experiments in the paper are also available.

• heuristicTime: int

  Execution time (in seconds) for TSPHeuristic. Default value is 20 seconds.

• plotUnorderedMatrix: boolean

  Export the unordered matrix as pdf. Default value is False.

• saveObjectiveFunctionValues: boolean

  Export the values of the objective functions as csv. Default value is False.

• printPermutations: boolean

  Print in the standard output the row and columns permutations. Default value is False.

• customColorSchemes: boolean

  Use the custom colour scheme. Default value is True.

Notes

If there is a clustering passed as an argument through rowClusters and columnClusters, then clusteringAlgorithm and nbClusters will not be used.

Hence, there are two ways to use the code:

• using clusteringAlgorithm and nbClusters to visualize the clustering obtained through the given clustering algorithm (see demo.py).
• using rowClusters and columnClusters to visualize the clustering given as an input files (see demoAdvanced.py).
• using rowClusters, columnCluster and inputIsList to visualize the clustering given as python lists (see demoAdvanced2.py).

Contact address: Thibault Marette: marette@kth.se