Skip to content

Latest commit

 

History

History
85 lines (51 loc) · 4.66 KB

Use GraphML in GraphOptima.md

File metadata and controls

85 lines (51 loc) · 4.66 KB

Use GraphML in GraphOptima

In this document, you will find:

  1. how to use your own graphml as the input for the GraphOptima.
  2. how to generate sample graphs similar to the default price network using graph-tool.

Note: To use your own graph, ensure that the graph has no parallel edges or self-loops. If the graph has parallel edges or self-loops, use the graph_cleaning.ipynb notebook to clean the graph. After that, it is recommended to reformat the cleared csv edgelist into a tsv (tab-separated values) file, since GraphOptima has been through tested with tsv files.

Prerequisites

  • Before proceeding, ensure the GraphOptima has been installed (refer to README.md).

Use Custom GraphML in GraphOptima

  1. Move your GraphML input graph to the input_graphs folder

    Make sure the GraphML file is located under the $SCRATCH/netviz/readability_optimization/input_graphs/graph_file_name.graphml. Change GRAPHML_FILE field of the readability_optimization/config.json to input_graphs/graph_file_name.graphml

  2. Remove redundant graph properties (Optional)

    A large GraphML file may slow down the framework and crush the layout generator. To avoid this, remove large text properties and those that are not used during the layout calculation. Check out readability_optimization/input_graphs/graph_cleaning.ipynb, which shows a example of using graph-tool to remove all the vertex and edge properties, parallel edges, and self-loops.

  3. Filter graph by GraphTrimmer (Optional)

    It the field of social network analysis, it is common to perform some pre-filtering of the graph to only focus on the interested relationships. Simpler graph by nature is easier to optimize and read. We provide a convent tool GraphTrimmer to filter the graph by the degree of the nodes. The GraphTrimmer is a Python script that can be found in the readability_optimization folder. The script will remove the nodes with degree less than the threshold. To use the GraphTrimmer, run the following command:

    python GraphTrimmer.py --input_graph input_graphs/graph_file_name.graphml --output_graph input_graphs/graph_file_name_trimmed.graphml --threshold 10 --filter_type reciprocal

    GraphTrimmer provides four modes of filtering: in, out, total and reciprocal.

    • in: filter the nodes with in-degree (a<-b) less than the threshold.
    • out: filter the nodes with out-degree (a->b) less than the threshold.
    • total: filter the nodes with total degree (in + out) less than the threshold.
    • reciprocal: filter the nodes with reciprocal (mutual, meaning a<->b) degree less than the threshold.

    NOTE: graph-tool is not available under Windows. To run the GraphTrimmer, it is recommended to use either the Singularity environment outlined in the README, or to use the following docker environment (since singularity is not available on Windows): https://hub.docker.com/repository/docker/0xc00005/netviz-graham-v10/general

How to generator sample graphs

  1. Enter the Singularity Environment

    Navigate to the appropriate directory and start the Singularity environment with the following command:

    module load singularity
cd $SCRATCH 
singularity shell --nv -B $SCRATCH $SCRATCH/netviz/readability_optimization/singularity/netviz-graham-v10.sif

  2. Init the Singularity Container

    Within the Singularity environment, set up the PATH variable and enable the RAPIDS environment with the following command:

    export MPLCONFIGDIR=$SCRATCH
source /opt/conda/etc/profile.d/conda.sh
conda activate base

  3. Navigate to input_graphs Folder

    Proceed to the sample graph folder.

    cd input_graphs

  4. Run graph-to-network Generation Script

    Execute the Python script generate_networks_with_gt.py . This generates various network types such as Price Network, Complete Graph, Block Model, and others.

  5. Run csv2tsv.py Script

    After the graph_tool_network_generation.py script finishes running, execute csv2tsv.py. This script first uses an open-source script graphml2csv.py provided by the Amazon Graph Science Team, graphml2csv.py, to convert GraphML files to CSV format. Then, it converts these CSV files to TSP.

    Upon completion, each GraphML network file hwill yield three different CSV/TSP files with distinct prefixes. The one used in our paper is fixed_price_10000_nodes_edges.csv and price_10000.graphml.

  6. Validate the Outputs

    Ensure that the graph with format "fixed_<graph_name>-edge.csv" is present in the sample_graphs directory. This file will be used as input for the GraphOptima framework.