16 Apr
2007
16 Apr
'07
7:17 p.m.
On 4/17/07, Fred lunnon <fred.lunnon@gmail.com> wrote:
Please elaborate on why you propose the error term (log N)/N. WFL
Ah yes, I see the point now: all strings of length up to n = log_2(N) should occur among the first N bits, so for example the count of the string "1" will not change while the sequence appends "0^n". Hence the best we can expect is that the relative frequency varies by log(N)/N. WFL