Open MikaelSlevinsky opened 6 years ago
@MikaelSlevinsky ! Welcome and thanks for bringing this up. I don't think that I can clarify this, because assuming that less floating point operations leading to a more precise result is wrong in the general case, as you have suitably pointed it out. I will strike that from the README.md. Thanks!
In the context of numerical evaluation of
L_{12}(1/2)
, the readme statesIs this in reference to the method named
goal12
, whose result appears to be-0.23164963886602852
?The evaluation of
L_{12}(1/2)
can be done inRational{BigInt}
that avoids floating-point rounding errors. The result is:Perhaps someone may clarify to me how the
goal12
method is more accurate than the other methods, notably theApproxFun
implementation that results in-0.2316496388655192
, which is two units of least precision away from the correctly rounded rational result.