A loss function 
assigns a numerical cost to taking action 
when the unknown state is 
. The risk of a decision rule is the expected
value of the loss under the sampling distribution, and a Bayes
decision rule minimizes the posterior expected
value of the loss. In classification, zero-one
loss assigns loss 0 to a correct classification
and loss 1 to an incorrect one. Squared-error and absolute
error loss are common for point estimation.
Loss Function
See also
Bayes Decision Rule, Classification, Decision TheoryExplore with Wolfram|Alpha
References
Berger, J. O. Statistical Decision Theory and Bayesian Analysis, 2nd ed. New York: Springer-Verlag, 1985.Cite this as:
Weisstein, Eric W. "Loss Function." From MathWorld--A Wolfram Resource. https://mathworld.wolfram.com/LossFunction.html