File:My Best Error.png

English: The complex error function has been on of the hardest functions I've tried to implement for my math library. There exist efficient algorithms for computing the real valued error function, so I naively thought this would be simple to extend to the complex plain. First, I tried using the tailor series, but this only proved accurate for points near the origin. So I tried using a different approximation for points far away, but regardless of where I chose the cutoff, I could still see tearing in the final image. Finally I tried interpolating between the two functions using a logistic curve, and that produced this. You can still see some tearing at the top and bottom, but the image is much smoother. Still, this is only a rough approximation of the true error function. I will probably replace it with something better if I ever get the determination to work on it again.