Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Excluding the Target Domain Improves Extrapolation: Deconfounded Hierarchical Physics Constraints

About

Extrapolation to out-of-distribution conditions is a fundamental challenge for physics-constrained deep generative models. Existing methods apply physical constraints as a single static regularization term uniformly across the generation process, and address neither the hierarchical structure of physical laws and the confounding variable problem. We propose the Deconfounded Hierarchical Gate (DHG), which serves as a diagnostic and control mechanism: it identifies when and how strongly temperature confounding contaminates each constraint level, so that hierarchical gates reflect intrinsic physical inconsistency rather than spurious temperature effects. DHG combines counterfactual estimation via the do-operator with backdoor adjustment to remove confounding, then applies Coarse-to-Fine physical constraints progressively. We report a counter-intuitive finding in pretraining: excluding the target-domain data from pretraining outperforms including it by 39% in extrapolation performance (RMSE 0.224 vs. 0.324). This occurs because FNO learns domain-agnostic physical patterns that transfer more effectively when the target domain is withheld. On a lithium-ion battery temperature extrapolation benchmark (trained at 24 degrees Celsius, evaluated at 4.0--43.0 degrees Celsius), our method achieves RMSE = 0.215, a 46% improvement over the unconstrained baseline (Pure CFM: 0.397).

Tsuyoshi Okita• 2026

Related benchmarks

TaskDatasetResultRank
Capacity scalar temperature extrapolationNASA Battery Dataset (B0005-B0036) (Near)
RMSE0.082
6
Voltage waveform temperature extrapolationNASA battery LiCoO2/graphite, 18650 cylindrical cells (Mean)
RMSE0.215
6
Voltage waveform temperature extrapolationNASA battery dataset Near LiCoO2 graphite 18650 cylindrical cells
RMSE0.259
5
Voltage waveform temperature extrapolationNASA battery dataset LiCoO2/graphite, 18650 cylindrical cells (Low1)
RMSE0.205
5
Voltage waveform temperature extrapolationNASA battery dataset LiCoO2/graphite, 18650 cylindrical cells (Low2)
RMSE0.217
5
Voltage waveform temperature extrapolationNASA battery dataset LiCoO2/graphite, 18650 cylindrical cells (High1)
RMSE0.229
5
Voltage waveform temperature extrapolationNASA battery dataset LiCoO2/graphite, 18650 cylindrical cells (High2)
RMSE0.178
5
Capacity scalar temperature extrapolationNASA Battery Dataset (B0005-B0036) (Low1)
RMSE0.085
3
Capacity scalar temperature extrapolationNASA Battery Dataset B0005-B0036 (Mean)
RMSE0.084
3
Showing 9 of 9 rows

Other info

Follow for update