Hi Mixumus,I just finished some research on the calibration of the Heston model. I tried several optimizers, including ASA. Some advice:- Check your error function and your option price calculation algorithm (I suggest using the version described in [The Litlle Heston Trap] as the version that does not give discontinuities/problem in taking logarithm of complex numbers).- Tweak the ASA-parameters: for instance the temperature_ratio_scale is quite influential in the optimization process- In general: it may very well be the case that there exist roughly equally well solutions to the calibration problem.Currently I perform calibrations over an error function on implied volatilities instead of over option prices. Additional weights are included to fit somewhat better to ATM options, and to lessen the effect of having a high concentration of options in some area in strike-maturity space. I calculate the option prices using the semi-closed solution for Heston, instead of FFT of the option price. I have found that these choices drastically change the optimization problem; in fact: I can use a MATLAB local optimizer, instead of some global optimizers I tested (among which was ASA). The local optimizer gives the calibration result no matter whatever initial guess I supply (obviously computation time differs ).Some other results I have found with the Heston calibration:Don't fit the Heston model to short-term options; the model is unable to account for jump-risks and cannot incorporate enough skewness in implied volatility for short-term. If short-term options are included, the calibration results will be bad for every maturity, since it tries to fit also the short-term option. I suggest using the Heston model only for options with maturity larger than 1 year. You might want to consider the value of 2*kappa*theta-sigma^2 (which should be larger than zero to avoid negative variance). During time the Heston paramters change somewhat, but this value is generally more stable.Good luck!

Hi. I'm the author of ASA, available at no charge from several locations:http://www.ingber.comhttp://alumni.calt ... ingSection 3.4. Use of Documentation for Tuning in ASA-README.[] (several formats) explains the necessity for tuning in all importance-sampling algorithms that are reasonably robust for many nonlinear systems. I think this addresses the main points brought up in this posting.ASA was used for options, to fit generic distribution to options data, e.g., a bottom-up approach to fitting strikes, instead of the usual top-down "smile" approach: %A L. Ingber %A C. Chen %A R.P. Mondescu %A D. Muzzall %A M. Renedo %T Probability tree algorithm for general diffusion processes %J Physical Review E %V 64 %N 5 %P 056702-056707 %D 2001 %O URL http://www.ingber.com/path01_pathtree.pdfASA is used in a full trading system that includes copula risk-management (including its good, and and ugly features): %A L. Ingber %T Trading in Risk Dimensions (TRD) %R Report 2005:TRD %D 2005 %I Lester Ingber Research %C Ashland, OR %O URL http://www.ingber.com/markets05_trd.pdfA shorter updated paper with this title appears in the Handbook of Trading: Strategies for Navigating and Profiting from Currency, Bond, and Stock Markets published by McGraw-Hill (2010). Only the Table of Contents and the Abstract appear in http://www.ingber.com/markets05_trd.pdf, as the publisher has requested all previous markets05_trd_report.pdf papers be removed from the internet. (Not online. Contact me for more information.)Lester