Euclidean mirrors and first-order changepoints in network time series
About
We describe a model for a network time series whose evolution is governed by an underlying stochastic process, known as the latent position process, in which network evolution can be represented in Euclidean space by a curve, called the Euclidean mirror. We define the notion of a first-order changepoint for a time series of networks, and construct a family of latent position process networks with underlying first-order changepoints. We prove that a spectral estimate of the associated Euclidean mirror localizes these changepoints, even when the graph distribution evolves continuously, but at a rate that changes. Simulated and real data examples on organoid networks show that this localization captures empirically significant shifts in network evolution.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Change Point Detection | Synthetic SBM datasets changing speed (test) | Hausdorff Distance12.96 | 49 | |
| Change Point Detection | Synthetic SBM changing speed CPD (full evolution) | F1 Score61.35 | 49 | |
| Change Point Detection | Synthetic SBM pace changes on fixed trajectories | Rand Index72.88 | 49 | |
| Change Point Detection | Synthetic SBM endpoint deviation and partial evolution | F1 Score49.3 | 49 | |
| Change Point Detection | Synthetic SBM endpoint deviation and full evolution | F1 Score52.86 | 49 | |
| Change Point Detection | Synthetic SBM Full evolution to endpoint graph | Rand Index66.42 | 49 | |
| Change Point Detection | CPD synthetic SBM endpoint deviation and partial evolution | Hausdorff distance21.56 | 49 | |
| Change Point Detection | Synthetic SBM Change with possibly partial evolution | Rand Index64.9 | 49 | |
| Change Point Detection | CPD with endpoint deviation and full evolution (SBM) | Hausdorff Distance21.7 | 49 |