XFlow Homepage | XFlow Paper | Documentation | Paper Collection

XFlow is a library built upon Python to easily write and train method for a wide range of applications related to graph flow problems. XFlow is organized task-wise, which provide datasets benchmarks, baselines and auxiliary implementation.

Update: FlowGPT: a custom GPT for graph dynamics analysis.

[comment]: <> (add icons https://css-tricks.com/adding-custom-github-badges-to-your-repo/)

Installation

pip install xflow-net

Example

Import XFlow

python import xflow.dataset.nx as nx_datasets import xflow.dataset.pyg as pyg_datasets import xflow.diffusion as diffusion_models import xflow.seed as seeds import xflow.util as util import xflow.method.im as im_methods import xflow.method.ibm as ibm_methods import xflow.method.cosasi.source_inference.multiple_source as source_inference

Influence Maximization

```python

Graphs to test

fn = lambda: nxdatasets.connSW(n=1000, beta=0.1) fn.name_ = 'connSW' gs = [fn, pyg_datasets.Cora]

Diffusion models to test

df = [diffusionmodels.SI, diffusionmodels.IC, diffusion_models.LT]

Seed configurations to test

se = [seeds.random, seeds.degree, seeds.eigen]

Configurations of IM experiments

imexperiments = [immethods.pi, immethods.eigen] rt = util.run( graph=gs, diffusion=df, seeds=se, method=imexperiments, eval='im', epoch=10, budget=10, output=['animation', 'csv', 'fig'] ) ```

Maximizing Blocking

python ibm_experiments = [ibm_methods.sigma, ibm_methods.degree] rt = util.run( graph=gs, diffusion=df, seeds=se, method=ibm_experiments, eval='ibm', epoch=10, budget=10, output=['animation', 'csv', 'fig'] )

See more examples in folder examples

Benchmark Task

Influence Maximization

• simulation: greedy, CELF, and CELF++,
• proxy: pi, sigma, degree, and eigen-centrality
• sketch: RIS <!-- - , SKIM, IMM -->

Blocking Maximization

• greedy
• pi
• sigma
• eigen-centrality
• degree

Source Localization

• NETSLEUTH (Legacy and Fast versions)
• Jordan Centrality
• LISN

Experimental Configurations

• Graphs: Compatible with graph objects/class by Networkx and Pytorch Geometric
• Diffusion Models: Support NDLib

How to Cite

We acknowledge the importance of good software to support research, and we note that research becomes more valuable when it is communicated effectively. To To demonstrate the value of XFlow, we ask that you cite XFlow in your work.

latex @article{zhang2023xflow, title={XFlow: Benchmarking Flow Behaviors over Graphs}, author={Zhang, Zijian and Zhang, Zonghan and Chen, Zhiqian}, journal={arXiv preprint arXiv:2308.03819}, year={2023} }

Contact

Feel free to email us if you wish your work to be listed in this repo. If you notice anything unexpected, please open an issue and let us know. If you have any questions or are missing a specific feature, feel free to discuss them with us. We are motivated to constantly make XFlow even better.