Halton sequence with QuantLib

[Carlo Vicentini]

Hello,

 

I would like to read your comments on some code I have written with QuantLib in mind. It deals with the Halton sequence and it is attached to this message.

 

I think the main problem with quasi-random numbers is how to take account of dimensionality, given a general Monte Carlo framework which does not (at least for the moment). I mean, different time steps should use Halton numbers generated from different "dimensions".

 

One way would be to use the parameter "seed" as a "proxy" for the dimension (this is the approach I followed). Another approach would be to modify the class interface so as to directly take account of the dimension.

 

In the attached files, "HaltonLds.*" contains the generator, and "qmceuropean.*" is a pricer adapted from "mceuropean"; "rngtypedefs.hpp" contains just a couple of new lines, and "qmcEuropeanOption.cpp" is an example adapted from "EuropeanOption.cpp".

 

I haven't tested the implementation thoroughly, yet, but at least the european option example seems to be correct.

Please tell me if I am missing something.

 

Regards,

 

Carlo Vicentini