diffuse: Diffuseness around centre of mass#
- hypercoil.functional.cmass.diffuse(X: Tensor, coor: Tensor, norm: Any = 2, floor: float = 0, radius: float | None = None) Tensor[source]#
Compute a compactness score for a weight.
The compactness is defined as
\(\mathbf{1}^\intercal\left(A \circ \left\|C - \frac{AC}{A\mathbf{1}} \right\|_{cols} \right)\mathbf{1}\)
- Dimension:
- Input : \((*, W, L)\)
*denotes any number of preceding dimensions, W denotes number of weights (e.g., regions of an atlas), and L denotes number of locations (e.g., voxels).- coor : \((*, D, L)\)
D denotes the dimension of the embedding space of the locations.
- Parameters:
- XTensor
Weight for which the compactness score is to be computed.
- coorTensor
Coordinates corresponding to each column (location/voxel) in X.
- norm
Indicator of the type of norm to use for the distance function.
- floorfloat (default 0)
Any points closer to the centre of mass than the floor are assigned a compactness score of 0.
- radiusfloat or None (default None)
If this is not None, then the centre of mass and distances are computed on a sphere with the specified radius.
- Returns:
- float
Measure of each weight’s compactness about its centre of mass.