This is exactly the same as above, except this time we minimize the distance between Z and Y using the Kullback norm (or something close to it). The regularized signal is called mreglog_signal#. The advantages of this distance are that the computations are much simpler (the numerical minimization is replaced by an analytical one), and that it allows for further generalizations. For instance, the forthcoming version of Fraclab will include a multiplicative transform of the exponents instead of a shift, in the case where this distance is use. Typical values for the shift in this case are around 1.