"scale" parameter value modified?

[eldarkurtic]

The following code with scale=0.01

x = NormalVariable(loc=0., scale=0.01, name="mu")
model = ProbabilisticModel([x])
sample = model.get_sample(2)
print(sample)

returns this:

   mu_scale  mu_loc        mu
0 -4.600166     0.0 -0.030332
1 -4.600166     0.0 -0.008252

Value of the mu_scale has been changed. Is this expected behavior since I have declared the random variable x with the scale value scale=0.01, but got something like log(0.01) = -4.600166.

[LucaAmbrogioni]

Hi cijad. Brancher automatically apply an inverse sofplus transformation when you initialize a strictly positive parameter. Then a sofplus transformation is added to the link. This allows to optimize everything using SGD. What the sample is returning is the pre-inverse softplus value instead of the scale itself. This is indeed missleading. The sample post-processing should re-apply the softplus. I will introduce this in the next release.