Animating the ForceAtlas2 algorithm

Straight edges	Curved edges

This script leverages a modification of the fa2 Python package, which allows the historical positions of the nodes (while the algorithm is iterating) to be returned. The (NetworkX) graph at each stage can then simply be visualised, and therefore animated.

Dependencies

• You will need to first clone the branch of my fa2 PR by the following command: git clone https://github.com/beyondbeneath/forceatlas2.git -b maintain-position-history /path/to/fa2-anim

• If you wish to plot aesthetic Bezier edges, you should copy the function available in the bezier-curved-edges-networkx repo.

• NetworkX: networkx==2.2

Sample usage

For this example, I will re-use the Game of Thrones network I gave in the bezier-curved-edges-networkx repo, but this will run on any NetworkX graph, represented here by the object G.

All we do here is setup a ForceAtlas2 function with appropriate parameters, and then run the animate_fa2 function:

f = ForceAtlas2(seed=100,
                outboundAttractionDistribution=True,
                edgeWeightInfluence=0,
                gravity=1,
                scalingRatio=10,
                verbose=False)

animate_fa2(G, f, num_iterations=100, output_dir='/tmp/fa2', edge_type='straight')

This will output num_iterations frames to /tmp/fa2. If you choose anything other than edge_type='straight' it will attempt to use the curved edges function that you will need to have in the same file. All the node & edge formatting are hard-coded in the function, but can be easily modified.

From here, you can combine the frames into an animated gif using your favourite tool. For example, using ImageMagick convert tool on MacOS (which can be installed using brew install imagemagick):

convert -delay 10 anim_frame_* animation.gif