TFNO 在 The Well 数据集上的超参调优结果显示基线被低估
Performance of Neural Operator Baselines on 'The Well' Dataset
The Well 数据集的官方基线其实没怎么调参，这篇只在 12 小时内调了下 TFNO，17 个任务里 15 个就大幅超过原基线，引用时得留心。
The Well 是一个 15TB、覆盖 16 个物理模拟领域的数据集，用作评估物理系统机器学习代理模型的基准。该数据集随附的神经算子基线原本只是低算力、几乎未调参的示例，却被后续工作当作完全调优的性能基准。研究者在同样 12 小时单次运行限制下，仅对 Tucker-factorized Fourier Neural Operator（TFNO）做了简短超参搜索，并只用单个输入时间步（原基线最多用 4 个历史步）。调参后，TFNO 在 17 个任务中的 15 个上 VRMSE 低于已发布的 FNO 和 TFNO 基线，有时低数个数量级。
Performance of Neural Operator Baselines on 'The Well' Dataset
The Well, a 15-terabyte dataset spanning 16 physical simulation domains, provides an extensive benchmark for developing and evaluating machine learning surrogate models for complex physical systems. The baselines released with the benchmark were presented by their authors as illustrative examples obtained at small computational cost, trained with zero to very limited hyperparameter tuning and under a highly constraining 12-hour per-run wall-clock limit. These results have subsequently been treated as fully tuned benchmarks, and in some cases effectively interpreted as unconstrained performance baselines. We perform a short non-exhaustive hyperparameter search for the Tucker-factorized Fourier Neural Operator (TFNO) under the same 12-hour per-run wall-clock limit, while using only a single input timestep rather than the history of up to four used by the published baselines. Despite this more restrictive setup, tuning alone reduces VRMSE below both published FNO and TFNO baselines on 15 of 17 tasks, sometimes by several orders of magnitude. This sensitivity to a modest tuning effort underscores that the released neural operator baselines should not be treated as architectural benchmarks.