edges.load_whitened_pcs

edges.load_whitened_pcs(
    volume_file,
    atlas=None,
    exclude_acronyms=DEFAULT_EXCLUDE_ACRONYMS,
    variance_threshold=DEFAULT_VARIANCE_THRESHOLD,
)

Whitened PCA of an encoding volume.

PCs are computed on the voxels inside the atlas mask, outside the excluded regions and with data (all-zero voxels are the volume’s “no probe coverage” sentinel). Each PC is signed so that its largest-magnitude loading is positive.

Parameters

Name	Type	Description	Default
volume_file	str or Path	Encoding volume .npz (brainwide_ephys_atlas_{res_um}um.npz).	required
atlas	iblatlas.atlas.BrainAtlas	Atlas at the volume’s resolution. Defaults to AllenAtlas(res_um).	None
exclude_acronyms	iterable of str	Regions left out of the PCA and of the valid mask.	DEFAULT_EXCLUDE_ACRONYMS
variance_threshold	float	Fraction of the variance the retained PCs must explain.	DEFAULT_VARIANCE_THRESHOLD

Returns

Name	Type	Description
	dict	pcs : (AP, ML, DV, k) float32 whitened PCs, NaN outside valid valid : (AP, ML, DV) bool mask of the voxels used explained : (k,) explained variance ratio loadings : pandas.DataFrame (k, n_features) feature_names : (n_features,) array project : callable mapping raw (n, n_features) values to whitened PCs res_um : int