MaskNearEmpty#
- class sdm.prior.tabular.transform.MaskNearEmpty(*, columns: tuple[float, float] = (0.5, 0.8), observed: HP[float] | None = None, min_observed: int = 4)#
Bases:
PriorTransformMask column groups to sparse, overlapping sets of observed rows.
- Parameters:
columns (tuple[float, float]) – Range for the fraction of feature columns to mask.
observed (HP[float] | None) – Sampled fraction of rows retained per column group. Groups in one dataset share a correlated rate. Defaults to
LogUniform(0.01, 0.1).min_observed (int) – Minimum retained rows per column group, capped by the row count.