My basic implementations of some of the transcendental functions I used in the fitting algorithm could produce inf or nan for certain inputs. This ultimately led to infinite loops in the fitting routine. I've copied over implementations from scipy source code that have better numerical properties.
My basic implementations of some of the transcendental functions I used in the fitting algorithm could produce inf or nan for certain inputs. This ultimately led to infinite loops in the fitting routine. I've copied over implementations from scipy source code that have better numerical properties.