Re: [Quantlib-dev] question on discountcurve fitting method

[Bojan N. <bo...@bn...>]

"Han, Guowen" <GH...@dt...> writes:

> Anyone know if there is any  particular reason for using Simplex method in FittedBondDiscountCurve class?


Probably a combination of:

- Simplex does not require derivatives (the Minpack LevenbergMarquardt
  used in QuantLib does not require them either but they are estimated
  internally by finite differences)

- It is slightly more robust to local minima compared to minimisation
  algorithms based on a current and trial point

Best,
Bojan


-- 
Bojan Nikolic          ||          http://www.bnikolic.co.uk