Max Mask Pixels

Optional upper limit on the number of active pixels in the discovered spine mask.

The spine mask is normally sized by the noise itself: pixels enter only if their correlation is statistically certain (Spine Threshold) and large enough to matter (Correlation Floor). Max Mask Pixels adds a hard cap on top of that.

How it works

• If the discovered spine exceeds the cap, the weakest pixels (lowest |correlation|) are dropped until the cap is met

• The mask stays a connected, 180-degree symmetric spine

• With no cap set, the mask size is determined entirely by the thresholds

When to use

• To bound training cost when the noise has very long-range correlation and the spine grows large

• To keep the mask compact if you find large masks remove too much context

Recommendations

• Leave unset (package default) for most data

• Prefer raising the Correlation Floor over capping: the floor removes pixels by relevance, while the cap is a blunt limit

• Watch the mask coverage (mask_rho2) when capping; a large drop means the cap is cutting into meaningful correlations