As to the comments, I have more sympathy with the point of view that in most cases comments are sign of the fact that code should be refactored.
As for me, in this example with optimal (from the author’s point of view) number of comments some of them are obviously needless:

// For the correct functioning of this algorithm
//quantity of massive elements must be equal to
// the quantity of elements in relative probabilities massive
This should be replaced with a set of assertions in the checkout version of a program, ASSERT etc.

// Notice: we don’t use $idx because
// we don’t need access to the single elements
// of the massive weights
This comment is relatively useless (although in this concrete case this is a rather doubtful question)… An experienced programmer won’t need it to realize how this all works. An inexperienced one will have to examine code anyway.