gcn-data-alignment

This is a Python implementation of subspace alignment measure (SAM) for quantitying the alignment of graph and features in deep learning, as described in our paper:

Yifan Qian, Paul Expert, Tom Rieu, Pietro Panzarasa, and Mauricio Barahona (2021), Quantifying the alignment of graph and features in deep learning, IEEE Transactions on Neural Networks and Learning Systems.

Installation

python setup.py install

Run the demo

cd alignment
jupyter notebook demo.ipynb

How to use the code

from alignment.optimizations import optimize_dim_subspaces

optimize_dim_subspaces(
    dataset="constructive_example",
    num_rdm=2,
    num_k=5,
    num_scanning=1,
    norm_type="Frobenius-Norm",
    log=False,
    heatmap=True
)

Cite

Please cite our paper if you use this code in your own work:

@article{qian2021quantifying,
  title={Quantifying the alignment of graph and features in deep learning},
  author={Qian, Yifan and Expert, Paul and Rieu, Tom and Panzarasa, Pietro and Barahona, Mauricio},
  journal={IEEE Transactions on Neural Networks and Learning Systems},
  year={2021},
  doi={10.1109/TNNLS.2020.3043196},
  publisher={IEEE}
}