System-Anchored Knee Estimation for Low-Cost Context Window Selection in PDE Forecasting

arXiv cs.AI / 3/27/2026

📰 NewsSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper identifies low-cost context-window selection as a distinct problem for fixed-window autoregressive neural PDE simulators, where existing methods can be expensive, brittle, or misaligned with rollout performance.
It proposes System-Anchored Knee Estimation (SAKE), a two-stage approach that first builds a small candidate set using physically interpretable system anchors and then selects the context window using knee-aware downstream criteria.
Experiments on eight PDEBench PDEBench families under a shared L∈{1,…,16} protocol show SAKE is the strongest overall matched-budget low-cost selector among evaluated baselines.
Reported results include 67.8% exact matches, 91.7% within-1 selection accuracy, 6.1% mean regret@knee, and a very low cost ratio (0.051) yielding 94.9% normalized search-cost savings.

Abstract

Autoregressive neural PDE simulators predict the evolution of physical fields one step at a time from a finite history, but low-cost context-window selection for such simulators remains an unformalized problem. Existing approaches to context-window selection in time-series forecasting include exhaustive validation, direct low-cost search, and system-theoretic memory estimation, but they are either expensive, brittle, or not directly aligned with downstream rollout performance. We formalize explicit context-window selection for fixed-window autoregressive neural PDE simulators as an independent low-cost algorithmic problem, and propose \textbf{System-Anchored Knee Estimation (SAKE)}, a two-stage method that first identifies a small structured candidate set from physically interpretable system anchors and then performs knee-aware downstream selection within it. Across all eight PDEBench families evaluated under the shared $L\in\{1,\dots,16\}$ protocol, SAKE is the strongest overall matched-budget low-cost selector among the evaluated methods, achieving 67.8\% Exact, 91.7\% Within-1, 6.1\% mean regret@knee, and a cost ratio of 0.051 (94.9\% normalized search-cost savings).