examples.complementarity.py

Complementarity Analysis for Graph Datasets using RINGS Framework

This script demonstrates how to: 1. Load a graph dataset from PyTorch Geometric 2. Apply various RINGS perturbations to the graph data 3. Compute complementarity metrics

Usage (from root directory):

python -m examples.complementarity –dataset MUTAG –perturbation original

For more options:

python -m examples.complementarity –help

examples.complementarity.get_available_perturbations()

Return a list of available perturbation names with descriptions.

Returns:

Mapping from perturbation name to description

Return type:

dict

examples.complementarity.create_perturbation(name, seed=42)

Create a perturbation transform based on the given name and seed.

Parameters:

• name (str) – Name of the perturbation

• seed (int) – Random seed for reproducible perturbations

Returns:

A transform that can be applied to graph data

Return type:

BaseTransform

examples.complementarity.apply_perturbation(dataset, perturbation)

Apply a perturbation to all graphs in a dataset.

Parameters:

• dataset – PyTorch Geometric dataset

• perturbation – Transform to apply

Returns:

List of transformed graph data objects

Return type:

list

examples.complementarity.create_complementarity_functor(n_jobs=1)

Create a ComplementarityFunctor with standard parameters.

The ComplementarityFunctor measures the difference between: 1. The metric space of node features (using euclidean distance) 2. The metric space of graph structure (using shortest path distance)

Parameters:

n_jobs (int) – Number of parallel jobs (-1 for all cores)

Returns:

Configured functor

Return type:

ComplementarityFunctor

examples.complementarity.compute_complementarity(dataloader, functor)

Compute complementarity scores for all graphs in the dataloader.

Parameters:

• dataloader – PyTorch Geometric dataloader containing graphs

• functor – ComplementarityFunctor to compute scores

Returns:

Array of complementarity scores

Return type:

numpy.ndarray

examples.complementarity.main()

Main function to run the complementarity analysis.

User inputs are handled via command line arguments:

–perturbation: “Perturbation to apply to the dataset (default: original)” –dataset: “Name of the TU dataset to use (default: MUTAG)” –seed: “Random seed for reproducibility (default: 42)” –batch-size: “Batch size for dataloader (default: 32)” –n-jobs: “Number of parallel jobs (-1 for all cores, default: 1)”