TOPICS
Search

Prophet Inequality


A prophet inequality compares the expected reward of an online selection rule with the expected best reward obtainable after all random values are known. In the classical one-choice problem, independent nonnegative integrable random variables V_1, ..., V_n arrive sequentially. The online rule may stop once and receives the current value, whereas the prophet receives max_(i)V_i. A factor-c guarantee means

 E[online reward]>=cE[max_(i)V_i].

The permitted selections, available distributional information, and arrival order are part of the model. The matroid prophet inequality allows selections forming an independent set of a matroid.


See also

Expectation Value, Matroid Prophet Inequality, Online Algorithm, Optimal Stopping

Explore with Wolfram|Alpha

References

Kleinberg, R. and Weinberg, S. M. "Matroid Prophet Inequalities." In Proc. Forty-Fourth Annual ACM Symposium on Theory of Computing (Ed. H. J. Karloff and T. Pitassi). New York: ACM, pp. 123-136, 2012. https://doi.org/10.1145/2213977.2213991.

Cite this as:

Weisstein, Eric W. "Prophet Inequality." From MathWorld--A Wolfram Resource. https://mathworld.wolfram.com/ProphetInequality.html

Subject classifications