Acknowledgement — orqviz (Zapata Engineering, Apache-2.0)
===========================================================

The divi.viz module implements loss-landscape scans on Divi's program model and
execution stack. API design and defaults are informed by the orqviz package as
follows:

* PCA scans (scan_pca): layout aligned with orqviz's PCA workflow—
  sklearn.decomposition.PCA on sample rows, a rectangular grid in PCA score
  space, inverse_transform into full parameter space, score-axis limits from the
  projected sample cloud with configurable offset padding, and optional
  anchoring away from the sample mean. Upstream reference implementations:
  src/orqviz/pca/data_structures.py, src/orqviz/pca/scans.py.

* Line and plane scans (scan_1d, scan_2d): default scalar ranges and
  random-direction behavior follow the same usage patterns as orqviz line and
  plane scans (coefficients along direction vectors; orthogonal second axis in
  2D). Upstream reference: src/orqviz/scans/scans_1D.py,
  src/orqviz/scans/scans_2D.py.

Upstream repository: https://github.com/zapata-engineering/orqviz
