I was surprised that in chapter 8.1 when faced with the problem of making sign function differentiable the proposed solution was different than in chapter 7.6 where the same problem was tackled.
There a logistic function was propsed, here a hyperbolic tangens.
Both of them are introduced as a typical sigmoid function, which gives a de ja vu perception, as the author seems not to remember that the concept was already introduced, instead presenting it as something totally new. More over - this time a different example is proposed as "natural".
The reader is left with a puzzle: are these two actually different names for the same function? If not, then which one is better, and what are factors by which one can compare such functions?