The Peres extractor just uses more of the data that the von Neumann throws away.
If the two consecutive bits are not different in the von Neumann extractor, then the algorithm just throws those two bits away and ignores them. The Peres extractor takes them and saves them for later to be reused in a manner that still preserves the 50-50 randomness we desire.
So the Peres extractor is preferable when we want to extract form 50-50 random data, but takes more time to run.
Latest Answers