CM++ Usage

Basic Usage

python -m hm01.cm
  -i {input network}
  -e {existing clustering (optional)}
  -c {clustering algorithm}
  -t {connectivity threshold}
  -n {number of processors (optional)}
  -o {output file}

• -i Input Network: The input .tsv edgelist filename.
• -e Existing Clustering: The existing clustering filename. This is a .tsv with each line being 'node_id cluster_id' format. If this isn't provided, CM++ will run an initial clustering on its own.
• -c Clustering Algorithm: The clustering paradigm to be used by CM++. Can choose from leiden, leiden_mod, ikc, and external.
• -t Connectivity Threshold: The connectivity threshold that every cluster must have at least to be considered 'well-connected'. This threshold can be a static integer e.g. 2 or a linear combination of:
• log10: $log10(n)$ where $n$ is the number of nodes in the cluster
• k: the k-core value of the cluster
• mcd: the minimum degree value
• e.g. 1log10, 1log10+1mcd+2k.
• -n Number of Processors: The number of cores to run CM++ in parallel. This defaults to 4.
• -o Output File: The output clustering file path. This is a 'node_id cluster_id' .tsv.

For example, you can run:

python -m hm01.cm -i network.tsv -e clustering.tsv -c leiden -g 0.01 -t 1log10 -n 32 -o output.tsv

External Clusterers

If you want to use an external clustering algorithm, use the following command format:

python -m hm01.cm
  -i     {input network}
  -e     {existing clustering (optional)}
  -c     external
  -cargs {clustering arguments json}
  -cfile {clustering algorithm code}
  -t     {connectivity threshold}
  -n     {number of processors (optional)}
  -o     {output file}

• -cargs: Clustering Arguments JSON: This is a json file containing argument names being mapped to their values. If the custom clusterer doesn't take any arguments (e.g. Infomap). The json will look like the following {} Else, here is an example arguments json for Markov Clustering (MCL): json { "inflation": 1.8 }
• -cfile: Clustering Algorithm Code: Path to your Python wrapper for the clustering algorithm. See this wrapper for example: Infomap. For instructions on writing your own wrapper, see here.