Home

Local2Global Embedding

This package implements the embedding methods used in [1]. This package uses pytorch and pytorch-geometric and it is a good idea to install these packages first following their respective installation instructions. If these packages are not already available during setup, an attempt is made to install them via pip which may not always work as expected. Afterwards use

pip install git+https://github.com/LJeub/Local2Global_embedding@master

to install the package and other dependencies. The patch alignment algorithm used in [1] is implemented in the separate local2global package. Installing this package will also install the latest version of local2global.

Command-line interface

This package exposes a command-line interface. Use

python -m local2global_embedding.run.run --help

to see the available options. For plotting the results, see

python -m local2global_embedding.run.plot --help

Note however that this will take a while to run.

For details of the python api see the module reference.

References

[1] (1,2)

L. G. S. Jeub, G. Colavizza, X. Dong, M. Bazzi, M. Cucuringu (2021). Local2Global: Scaling global representation learning on graphs via local training. DLG-KDD’21. arXiv:2107.12224 [cs.LG]

Contents

• Reference
  • local2global_embedding

Index

• Index