Problem with cumulation for covariance matrix (pc) using B*D*z~N(0,C)

[Antoine948398]

Hello,

I use CMAES and its C implementation to deal with stress inversion.
When my problem is very ill conditionned, I have a floating point exception in cmaes.c at line 893/894

/* cumulation for covariance matrix (pc) using B*D*z~N(0,C) */                                                                                                                                      
 hsig = sqrt(psxps) / sqrt(1. - pow(1.-t->sp.cs, 2*t->gen)) / t->chiN 
  < 1.4 + 2./(N+1); 

its with the "pow" function, t->gen = 752 and 1.-t->sp.cs = 0.62435950746181024, so the result must be really close to 0....

Have you got an idea how to avoid this kind of problens, moreover, it has begin to converge !!

Cheers,

Antoine

[Antoine948398]

I have managed to solve the problem by using a "custom pow" function that avoid small values:

static double custom_pow(double a, double b)
{
  double i = log(a);
  if ( b*i < -10) return 0;
  return exp(b * i);
}

[nikohansen]

This must be a machine dependent problem, as t->gen becomes always large. Your solution seems spot on. I would guess it should also be possible to configure the compiler to ignore numerical underflow errors, which is relatively safe to do.