pycrostates.segmentation.auto_information_function#

pycrostates.segmentation.auto_information_function(segmentation, lags, ignore_repetitions=False, log_base=2, n_jobs=1)[source]#

Compute the Auto-information function (aif).

Compute the Auto-information function (aif) as described in Von Wegner[1]:

\[I(X_{n+k} ; X_{n}) = H(X_{n+k}) - H(X_{n+k} | X_{n}) = H(X_{n+k}) - H(X_{n+k},X_{n}) + H(X_{n}) = H(X_{n+k}) + H(X_{n}) - H(X_{n+k},X_{n})\]

Parameters:

segmentationRawSegmentation | EpochsSegmentation: Segmentation object containing the microstate symbolic sequence.
lagsint | list | tuple | array of shape (n_lags,): Lags at which to compute the auto-information function. If int, will use lags = np.arange(lags).
ignore_repetitionsbool: If True, ignores state repetitions. For example, the input sequence AAABBCCD will be transformed into ABCD before any calculation. This is equivalent to setting the duration of all states to 1 sample.
log_basefloat | str: The log base to use. If string: * bits: log_base = 2 * natural: log_base = np.e * dits: log_base = 10 Default to bits.
n_jobsint | None: The number of jobs to run in parallel. If -1, it is set to the number of CPU cores. Requires the joblib package. None (default) is a marker for ‘unset’ that will be interpreted as n_jobs=1 (sequential execution) unless the call is performed under a joblib.parallel_config context manager that sets another value for n_jobs.

Returns:

lagsarray of shape (n_lags,): Lag values in sample.
aiarray of shape (n_lags,): Time-lagged mutual information array for each lag.

References

Examples using `pycrostates.segmentation.auto_information_function`#