`synthesise_from_cov_and_spectrum`#

hypercoil.neuro.synth.synthesise_from_cov_and_spectrum(spectrum: Tensor, cov: Tensor, *, key: PRNGKey) → Tensor[source]#

Create a synthetic signal matched in spectrum and covariance to references.

The synthetic signal will be drawn from a stationary Gaussian process. After sampling i.i.d. from a standard normal distribution, the synthetic data are transformed to match the references.

Dimension:

spectrum : \((D, F)\): D denotes dimension of the multivariate signal. F denotes the frequency dimension of the signal.
cov : \((D, D)\): As above.
output : \(D, N\): D as above. N is the time dimension of the signal, i.e., 2 * (F - 1).

Parameters:

spectrumtensor: Spectrum to be matched.
covtensor: Covariance matrix to be matched.
dtypetorch.dtype (default None): Data type of the synthetic dataset.
devicetorch.device (default None): Device on which the synthetic dataset is instantiated.

See also

synthesise_matched()
match_cov_and_spectrum()