For example, it seems to determine de bins from the first N samples accumulated. what if you pass less elements than this. Are the bins completely fixed after that? What was the logic behind that design rather than, for example, to somehow declared uniform bins in a range?

What if I have to define uniform equispaced bins, do I pass a uniform sequence at the beginning? but then those elements will be counted?

I would like to define uniform bins too, but If i read the source correctly this is not possible at the moment. Maybe I will write a such a density estimator once.

How the counters or probabilities are accessed at the end? What else can be done with the density accumulator object, can the cummulative or the inverse pdf be accessed easily within the library. Can it be plugged in into a random number generator as a distribution function.

If values are found which lie outside the histogramm they are sorted into the first of the last bin.