Department of Pattern Recognition

(eng): An attempt to formalize heuristic concepts like strings (sequence resp.) "typical" for a probability measure is stated in the paper. Both generating and testing of such strings is considered. Kolmogorov complexity theory is used as a tool. It is shown that no pseudorandom generator can produce long "typical" strings. The time complexity of pseudorandom generators with oracles capable to recognize "typical" strings is shown to be at least exponential with respect to the length of the output.