Compare the simplified calibration strategy with official PETSA

Compare the paper's simplified per-channel affine input-calibration-plus-linear-output-head strategy directly with the official PETSA implementation under the same leakage-free online forecasting evaluation protocol.

Background

The paper evaluates a simplified calibration variant inspired by PETSA rather than the official PETSA implementation. This prevents the reported calibration-based accuracy, memory, and compute results from establishing how the simplified strategy compares with PETSA itself.

The authors explicitly defer that direct comparison to future work, leaving the relative effectiveness and resource trade-offs unresolved.

References

A direct comparison of our simplified calibration point with the official PETSA implementation is likewise left for future work.

— When Does Online Adaptation Pay on the Edge? A Leakage-Free Evaluation of Warmup, Learning-Rate Selection, and Resource Trade-offs for Time-Series Forecasting  (2609.01126 - Fujimoto et al., 1 Sep 2026) in Discussion and Limitations, Section 5