deploy.

2026-09-11 03:46:11 +08:00
parent 45b9f9ed5d
commit b7f55aa631
114 changed files with 3646 additions and 576 deletions
@@ -460,6 +460,14 @@ see <a href="../technique/measurement/training-cut-set-design.html">Designing a
 <p>After training, the system reports three quality metrics:</p>
 <h3 id="correlation-coefficient-r">Correlation Coefficient (R)</h3>
 <p>A single value for the overall result. Ranges from 0 to 1; for new tools, expect 0.95–0.999.</p>
+<p><strong>R measures waveform agreement after phase alignment, and does not rank the coefficients that
+produced it.</strong> A cut set that leaves a coefficient combination unobservable still reports R above
+0.95, over shear values orders of magnitude too large; and between two runs on the same
+data, the one whose coefficients are closest to the truth can be the one with the lower R, because
+an exact step-to-shot pairing window scores lower than a window wide enough to smear the waveform.
+Read R as a floor the data has to clear, then judge the run by the force it reproduces over time.
+What the set can determine at all is fixed before cutting — see
+<a href="../technique/measurement/training-cut-set-design.html">Designing a Training Cut Set</a>.</p>
 <h3 id="training-error-ratio-trainingerrratio">Training Error Ratio (TrainingErrRatio)</h3>
 <p>A per-step variable registered automatically after training. Lower values indicate better step-level quality:</p>
 <div class="math">