An urn model from learning theory
Stéphane Boucheron, Dani�le Gardy · Random Structures and Algorithms · 1997
We present an urn model that is a variation of the classical occupancy model, and in which the balls are of two types (good and bad). We analyze the number of urns that contain more (or fewer) good balls than bad balls. We find Gaussian limiting distributions in the static case and convergence of the finite-dimensional distributions towards those of a Gaussian, non-Markov process in the dynamic case. © 1997 John Wiley & Sons, Inc. Random Struct. Alg., 10, 43–67 (1997)