Serving the Quantitative Finance Community
Heston Calibration
Hi,I try to calibrate the heston model to equity implied volatility surface. I face to issues as: (1) selecting a good objective function, (2) Optimization method (GA, SA or others).Can you help me on above topics and others if you think that i must focus on others crucial points.Thanks a lot
Heston Calibration
I've seen that you haven't received any answer yet. Maybe that if we're two.. somebody would answer us.. 
I'm trying to calibrate the heston model too, on Eurostoxx 50 vanilla options. my pricer should be ok but I don't really know which method to use for the calibration: I've tried the SA in C# but the best result I've got is 4% in relative error.. not got enough for a calibration! Thus, I'd like to put constraints on my parameters so that the optimizer doesn't return me negative vols etc.. could someone advise me on it?
I'm trying to calibrate the heston model too, on Eurostoxx 50 vanilla options. my pricer should be ok but I don't really know which method to use for the calibration: I've tried the SA in C# but the best result I've got is 4% in relative error.. not got enough for a calibration! Thus, I'd like to put constraints on my parameters so that the optimizer doesn't return me negative vols etc.. could someone advise me on it? 
Heston Calibration
You should try out the Differential Evolution (DE) algorithm. It seems to be working great for me, and can successfully calibrate the Heston Model to a few implied volatility surfaces that I have tried.As a starting point, check out the DE website.If you need more information and explanation you should try to obtain a copy of the book on DE. In this book the algorithm is explained in detail. This includes an explanation on how to use the algorithm to minimize a function with parameter constraints. As you know, we have such constraints in the Heston Model.
-
- commoquant
- Posts: 0
- Joined:
Heston Calibration
Have you tried good ole hill climbing methods?LBFGS or Levenberg-Marquadt work fine for me.
Heston Calibration
QuoteOriginally posted by: commoquantHave you tried good ole hill climbing methods?LBFGS or Levenberg-Marquadt work fine for me.As for myself, I have not tried those (yet). I guess that those are much more likely to get stuck in local minima. Especially because one doesn't actually know at what point to start the local minimization algorithm. I have however heard of people running Differential Evolution for a little time and then running a local algorithm like Levenberg-Marquardt. This will increase the chances considerably of one finding the global minima.