docs/glod.md

**TODO**: summarize from <https://blog.derwen.ai/graph-levels-of-detail-ea4226abba55>

Overall, this approach relies on a notion of developing "abstraction layers" atop graph data: how can graphs be analyze at differing levels of detail?

For this one must understand patterns in graphs such as network motifs and how to run topological transforms to help identify patterns.
A side benefit is that such transform can help boost the invariance of graph representation used when training models.