1711
Progress Journals & Experimental Routines / Re: ADARQ's journal
« on: April 07, 2016, 06:02:53 pm »
Raptor I just saw this.
It is almost impossible indeed, when you throw away significant digits you can't ask them back later.
The only way to possibly make this work is if the prices are not random but specified and they also don't round/trunk to the same one.
Simplified example: you can only store integers. If your floats are random then it is impossible to tell 1.2 from 1.3, they both give 1 or 2. But if the floats ars specified and far apart, e.g 1.2 and 3.2, then it works because 1.2 gives 2, 3.2 gives 4. In other words, if the accepted significant digits are different for each number then you can successfully map, use the 'wrong' values for quantifying and then retrieve the correct/long ones for the exact calculation. See it as dictionary where the labels are the rounded values, if they are unique it works.
Get it? If not I'll come back with more detailed/tailored example for both cases.
It is almost impossible indeed, when you throw away significant digits you can't ask them back later.
The only way to possibly make this work is if the prices are not random but specified and they also don't round/trunk to the same one.
Simplified example: you can only store integers. If your floats are random then it is impossible to tell 1.2 from 1.3, they both give 1 or 2. But if the floats ars specified and far apart, e.g 1.2 and 3.2, then it works because 1.2 gives 2, 3.2 gives 4. In other words, if the accepted significant digits are different for each number then you can successfully map, use the 'wrong' values for quantifying and then retrieve the correct/long ones for the exact calculation. See it as dictionary where the labels are the rounded values, if they are unique it works.
Get it? If not I'll come back with more detailed/tailored example for both cases.