LFPCompressed

LFPCompressed(
    U_scaled,
    Vh_hat,
    ns_original,
    epsilon,
    alpha,
    cr_svd,
    cr_wp,
    cr_total,
    left_overlap=0,
    ns_extended=0,
)

Compressed representation of an (nc, ns) LFP matrix.

Attributes

Name	Type	Description
U_scaled	(ndarray(nc, r), float32)	Left singular vectors scaled by their singular values: U[:, :r] * sv[:r].
Vh_hat	(ndarray(r, n_wp_slots), float32)	WP-domain leaf-node coefficients (alpha > 0) or time-domain rows (alpha == 0).
ns_original	int	Output time samples after trimming guard bands.
epsilon	float	SVD threshold multiplier used during compression.
alpha	float	WP threshold multiplier used during compression.
cr_svd	float	SVD-only compression ratio ncns / (r(nc+ns)) — what CR would be without WP.
cr_wp	float	WP compression ratio r*ns / n_kept (1.0 when alpha=0).
cr_total	float	True compression ratio ncns / (rnc + n_kept): original floats / stored floats.
left_overlap	int	Guard-band samples to trim at left after inverse WP (0 when no overlap context).
ns_extended	int	Total time samples the WP coefficients represent (0 → use ns_original).