pyphi.measures.distribution.intrinsic_differentiation#

pyphi.measures.distribution.intrinsic_differentiation(p, state)[source]#

Smallest positive surprisal in a repertoire slice.

Selects p.squeeze()[state], takes its pointwise intrinsic differentiation (the surprisal \(-\log_2 p_i\); see pointwise_intrinsic_differentiation()), and returns the minimum over the entries that are positive up to config.numerics.precision. If no entry is positive (every selected probability equals 1 up to floating-point noise), the result is 0.0.

Parameters:

  • p (np.ndarray) – A probability distribution.

  • state (State) – Index into the squeezed distribution selecting the slice to score.

Returns:

DistanceResult – The minimum positive surprisal of the selected slice (of the whole repertoire when state is None), with method="INTRINSIC_DIFFERENTIATION".