23 Feb
2004
23 Feb
'04
8:38 p.m.
Date: Mon, 23 Feb 2004 13:30:11 -0500 From: Michael Kleber <kleber@brandeis.edu> "10" -> 0 "01" -> 1 throw away "11" and "00". This cuts your output rate in half (-2epsilon^2), but if there's really no short-term correlation, you get true randomness. That actually cuts your output rate by a factor of 4, not 2, right? If 2 input bits always gave you an output bit, that would cut the rate in half. But 2 input bits gives you an output bit only half the time, so it's a factor of 4. Andy