In ML we often have to store a huge amount of values ranging from 0 to 1, mostly being probabilities. The most common data structure to do so seems to be a floating point? Indeed, the range of floating points is huge. This makes them imprecise in the desired interval and inefficient, right?
This question suggests using the biggest integer value to represent a 1 and the smallest for 0. Also, it points to the Q number format, where all bits can be chosen as fractional, which sounds very efficient.
Why have these data types still not found their ways into numpy, tensorflow etc.? Am I missing something?