As long as you don't need the infrastructure with derivatives and parametrics that is associated with the original class, it should be possible to extract the relevant computation (which does not contain any real "state", but only depends on the three parameters that you mentioned) and place it into a single static method.
Based on the original Gaussian class, it should roughly look like this:
// Based on the Gaussian class from Apache Commons Math
public static double gaussian(double x, double mean, double sigma)
{
double norm = 1 / (sigma * Math.sqrt(2 * Math.PI));
double is = 1 / sigma;
double i2s2 = 0.5 * is * is;
double xMinusMean = x - mean;
return norm * Math.exp(-xMinusMean * xMinusMean * i2s2);
}