To make evolution videos #443, we need to determine vertex coordinates in each frame. This can be done using the spring-electrical simulator, where vertices are considered charged particles connected by springs corresponding to edges.
We need code that will take a single hypergraph history and produce a list of frames, each containing coordinates of all vertices.
Ideally, we want an n log n GPU implementation. But it will take a few steps to get there:
[ ] n^2 CPU implementation (C++ or compiled WL)
[ ] n log n CPU implementation (C++ or compiled WL)
[ ] n^2 GPU implementation
[ ] n log n GPU implementation
It would be nice to expose this as separate property in case users want to do something else with these coordinates or use a custom renderer.
Additional context
To create evolution videos, we will also need to figure out the self-loops' orientation, as we cannot allow it to jump around frame-to-frame. But that will be a separate issue.
The problem
To make evolution videos #443, we need to determine vertex coordinates in each frame. This can be done using the spring-electrical simulator, where vertices are considered charged particles connected by springs corresponding to edges.
We need code that will take a single hypergraph history and produce a list of frames, each containing coordinates of all vertices.
Possible solution
There is a partial implementation in the
evolutionVideobranch.Ideally, we want an n log n GPU implementation. But it will take a few steps to get there:
It would be nice to expose this as separate property in case users want to do something else with these coordinates or use a custom renderer.
Additional context
To create evolution videos, we will also need to figure out the self-loops' orientation, as we cannot allow it to jump around frame-to-frame. But that will be a separate issue.