compress_pipeline

compress_pipeline(
    data,
    h=None,
    epsilon=100.0,
    alpha=7.0,
    fs=250.0,
    cadzow_rank=None,
    cadzow_niter=None,
    cadzow_fmax=None,
)

Full LFP compression pipeline: Cadzow denoise → SVD-adapt → WP threshold.

Parameters

Name	Type	Description	Default
data	ndarray of shape (nc, ns), float32	Raw LFP data at fs Hz. Rows are channels, columns are time samples.	required
h	dict or None	Neuropixel probe header containing ‘x’ and ‘y’ channel coordinates. Defaults to the first nc channels of the NP1 (version 1) geometry. Pass neuropixel.trace_header(version=2) for NP2 probes.	None
epsilon	float	SVD-adapt threshold multiplier. Default 100.	100.0
alpha	float	WP threshold multiplier. Default 7.	7.0
fs	float	LFP sampling rate [Hz]. Default 250.	250.0
cadzow_rank	int or None	Maximum Cadzow rank. None uses the default for the geometry (5, see cadzow_lfp_kwargs).	None
cadzow_niter	int or None	Number of Cadzow iterations. None uses the default for the geometry (1).	None
cadzow_fmax	float or None	Maximum frequency passed to the Cadzow denoiser [Hz]. None processes all bins (the default for the geometry).	None

Returns

Name	Type	Description
reconstructed	ndarray of shape (nc, ns), float32	Denoised and compressed–reconstructed LFP.
compressed	LFPCompressed	Compression metadata and per-stage compression ratios.