If solution to a problem is equally sensitive to 2 inputs, X and Y, but X has twice the uncertainty level due to additive noise, how would you budget n samples between X and Y to get to balanced solution uncertainty?
Feels like a mix of Kalman filter and multiarmed bandit.
Conversation
Replying to
Problem is under defined - answer depends on the absolute value of the uncertainty U in Y that is being “doubled”.
1
Replying to
Yes, make assumptions as needed for underdefined parts. Can only put partial problem into a tweet size :)
Replying to
Probably you mean square-root-2 times the level of uncertainty?
1
1




