Re: PSLQ

*To*: mathgroup at smc.vnet.net*Subject*: [mg80860] Re: PSLQ*From*: Clashton at gmail.com*Date*: Tue, 4 Sep 2007 03:49:24 -0400 (EDT)*References*: <200709010435.AAA26690@smc.twtelecom.net>

Thanks. I have tried your code an it appears to do exactly what I want. Jimmy. On Sep 3, 6:13 am, d... at wolfram.com wrote: > I suspect that Bertok code is doing roughly as it should. In general PSLQ > just "runs out of precision", so to speak, and has to give up, when there > is no reasonable candidate relation. If you in fact want to see huge > multipliers, you can get that effect using lattice reduction methods. But > I suspect you just want a graceful failure mode. > > Around 9-10 years ago I wrote a PSLQ implementation that I think is a bit, > well, not so good. I give the code below, with the caveat that it is "as > is". It might or might not find your desired integer relations. Don't > expect it to find you amorous relations, or wealthy distant relations. I > simply offer it as an alternative to what you already have, on the off > chance that it might be of use. Feel free to improve on it in any way you > like (and if it starts finding wealthy distant relations, I'd like to hear > about that). > > I no longer recall how this stuff works. My only comment on the actual > code is that the value 'const' can be made into a parameter for this > algorithm; it is equivalent to lambda^2 from PSLQ literature. >