Dimensionality reduction

feedbax.dimred.pca (x: Shaped[Array, '*batch features'],**kwargs: Any) -> PCAResults

Principal component analysis.

Takes the last axis of x as features. Flattens the preceding dimensions, then mean subtracts the features and performs singular value decomposition. Unflattens the PCs back to the original shape before returning.

Parameters:

Name	Type	Description	Default
x	Shaped[Array, '*batch features']	Input features.	required
**kwargs	Any	Additional arguments for jax.numpy.linalg.svd.	{}

feedbax.dimred.PCAResults (Module)

Attributes:

Name	Type	Description
components	Array	Principal components.
singular_values	Array	Singular values \(S\).