To install

It's suggested to install new packages in a virtual environment. But this one is safe if you know what you are doing.

$ pip install -e .

EXAMPLES

To produce node embeddings of an input graph:

$ python -m forcedirected embed fdbasic --help 
$ python -m forcedirected embed fdshell --help
$ python -m forcedirected embed fdtargets --help
$ python -m forcedirected embed node2vec --help

Such as:

$ python -m forcedirected embed fdtargets --ndim 128 --edgelist ./data/graphs/cora/cora_edgelist.txt --epochs 1000 --name cora --verbosity 2

The string after embed, such as fdtargets and fdshell indicate the embedding method.

To generate synthetic graph:

$ python -m forcedirected generate --help
$ python -m forcedirected generate lfr --help 

Such as:

$ python -m forcedirected generate lfr -n 1000 # Use LFR benchmark algorhtm to generate graph with 1000 nodes

CODE USAGE

If you want to use the force-directed module in code, you could have a look at the embed_* functions in ./forcedirected/embed/fd.py.

The actualy force-directed class is in ./forcedirected/models. The file ForceDirected.py contains the base class which other classes must derive from. Other files include a class for each method. The main difference between methods are the criteria for picking node pairs for calculating the force between them. For example model_201_basic takes all available node pairs and calculates the forces in between with the same coefficients.

The implementation for [^fdshell,^fdproof] papers is model_204_shell. In [^fdproof] the emphasis was more on a mathematical proof.

[^fdshell] Lotfalizadeh, Hamidreza, and Mohammad Al Hasan. "Force-directed graph embedding with hops distance." 2023 IEEE International Conference on Big Data (BigData). IEEE, 2023.

[^fdproof] Lotfalizadeh, Hamidreza, and Mohammad Al Hasan. "Kinematic-Based Force-Directed Graph Embedding." International Conference on Complex Networks. Cham: Springer Nature Switzerland, 2024.