To simplify my task, let's just say that we have a situation when we have to choose a point and we have N=2:7 points to choose from. We have a criteria by which probabilities are built, but the difference of calculated probabilities is very weak.
For example, we got N = 3, and probabilities are [0.31 0.38 0.31]. (Some points may have equal probability to be chosen). Or, for another example, N = 5 and probabilities are
[0.2004 0.2011 0.1998 0.1999 0.1987]
What we need - is to amplify the difference of probabilities - to make the biggest probability very strong, and others - very weak.
For example:
F_amp([0.31 0.38 0.31]) returns [0.11 0.78 0.11],
or F_amp([0.2004 0.2011 0.1998 0.1999 0.1987]) returns
[0.2 0.5 0.12 0.13 0.05]
The question is - what function to use for this task? We can't change the physical meaning of criteria - our criteria is just as we have - we just have to exaggerate the fact that point with highest probability will be chosen with significantly larger chance.
P.S. Sorry, if I have chosen wrong stackexchange-forum for this question. Feel free to correct or move question if needed.