fnerrf2.gms : Test correctness of errorf intrinsic
this data set should be more precise, as precise as you can be with
Matlab.
we use two identities to compute errorf:
erf(-z) = -erf(z)
computed using errorf(x) = 0.5 * (1+erf(x/sqrt(2)));
Small Model of Type: GAMS Includes: fnset_x.inc fntest_x.inc
<![CDATA[
$title 'Test correctness of errorf intrinsic' (FNERRF2,SEQ=187)
$include fnset_x.inc
* this data set should be more precise, as precise as you can be with
* Matlab.
* we use two identities to compute errorf:
* erf(-z) = -erf(z)
* computed using errorf(x) = 0.5 * (1+erf(x/sqrt(2)));
$gdxin fnerrf2
$load aeps0 aeps1 aeps2 reps0 reps1 reps2 relToInput T data
$gdxin
reps1 = 2e-14;
loop {T,
data(T, 'f') = errorf.value (data(T,'x'));
data(T, 'fx') = errorf.grad (data(T,'x'));
data(T,'fxx') = errorf.hess (data(T,'x'));
data(T, 'rc') = mathlastrc;
data(T, 'ec') = mathlastec;
};
$include fntest_x.inc
]]></plaintext></body></div>