Closed sei-jklein closed 7 years ago
For each alternative, choose N cost values, sampled from the interval (input cost, 3x input cost). N is a small number like 3 or 5.
This gives us R = N * number of nodes.
Run the optimizer with all costs set to the input values. Repeat R times, changing one cost value each time. Cluster the results.
(This won't actually require R runs, since you can abort any run that skips the node that you are twiddling.)
This is similar to the Twiddle the Cost operator, but with direct cause-and-effect visibility.
bigfatnoob
commented
8 years ago
For each alternative, choose N cost values, sampled from the interval (input cost, 3x input cost). N is a small number like 3 or 5.
This gives us R = N * number of nodes.
Run the optimizer with all costs set to the input values. Repeat R times, changing one cost value each time. Cluster the results.
(This won't actually require R runs, since you can abort any run that skips the node that you are twiddling.)
This is similar to the Twiddle the Cost operator, but with direct cause-and-effect visibility.