Learning Transferable Predictability Representations

In three linesMethod to assign consistent predictability scores to short trajectory windows on a deterministic-stochastic continuum. GON (Gauge-Fixed Ordinal Network), a temporal convolutional model, resolves cross-system ambiguity via anchor-and-variance objective. Transfer validated on 5 dynamical systems, outperforming scratch training across all window budgets.

Summary generated by Claude — human-verified