26 Nov
2002
26 Nov
'02
8:18 p.m.
I wrote:
Has anyone proven that one can't do better with any method of approximation? I guess one could do it with a counting argument; heuristically I've only got lg n bits to spend. If I spend them all on one number, I can make it as small as I want (the penultimate convergent from the continued fraction gives the modular inverse of m), but if I split it among two of them, I can only get them half as small. -- Mike Stay staym@datawest.net