Strong supermartingales and limits of nonnegative martingales (Q272945): Difference between revisions

Loosely speaking, Komlós' lemma says that given an infinite bounded sequence of random variables there exists a random variable and a subsequence such that the Cesáro means of any subsequence converge almost surely to this random variable. The lemma holds also if one refers to convex combinations of subsequences. In the present paper, the authors switch from random variables to martingales, and they consider different versions of Komlós' lemma for non-negative martingales, which hold only with convergence in probability.