48282d57a0f11094d71c7310898ab347e6b847b3,nilearn/signal.py,,_standardize,#,23

Before Change

        signals = signals.copy()

    if normalize:
        if signals.shape[0] == 1:
            warnings.warn("Standardization of 3D signal has been requested but "
                "would lead to zero values. Skipping.")
            return signals

        if not detrend:
            // remove mean if not already detrended
            signals = signals - signals.mean(axis=0)

        std = np.sqrt((signals ** 2).sum(axis=0))
        std[std < np.finfo(np.float).eps] = 1.  // avoid numerical problems
        signals /= std
    return signals

After Change

        copy of signals, standardized.
    

    if standardize not in [True, False, "psc", "zscore"]:
        raise ValueError("{} is no valid standardize strategy."
                         .format(standardize))

    if detrend:
        signals = _detrend(signals, inplace=False)
    else:
        signals = signals.copy()

    if standardize:
        if signals.shape[0] == 1:
            warnings.warn("Standardization of 3D signal has been requested but"
                          " would lead to zero values. Skipping.")
            return signals

        elif (standardize == "zscore") or (standardize is True):
            if not detrend:
                // remove mean if not already detrended
                signals = signals - signals.mean(axis=0)

            std = signals.std(axis=0)
            std[std < np.finfo(np.float).eps] = 1.  // avoid numerical problems
            signals /= std

        elif standardize == "psc":
            mean_signal = signals.mean(axis=0)
            invalid_ix = mean_signal < np.finfo(np.float).eps
            signals = (signals / mean_signal) * 100
            signals -= 100

            if np.any(invalid_ix):
                warnings.warn("psc standardization strategy is meaningless "
                              "for features that have a mean of 0 or "
                              "less. These time series are set to 0.")
                signals[:, invalid_ix] = 0

    return signals


def _mean_of_squares(signals, n_batches=20):

Instances