[Quantlib-dev] Finite Difference with Boundary Fixing

[Kuan-Khoon T. <kt...@en...>]

Hi,

I am new both to c++ and QL.
Recently, I completed a project on solving the BS for Vanilla American 
put using boundary fixing, ie use S=s-p0, where s is stock price and p0 
is exercise boundary, thereby fixing 0<S<Inf. The resulting equations is 
non-linear (will be linearised for numerical purpose) and will have 
time/S dependent coeffs even for const sigma and r.

I did it using MATLAB, now I am learning c++ so I was thinking of 
implementing it in QL as a practice.

My initial framework are as follows:

Construct a new FDBFAmericanEngine class, derived in much the same way 
as how FDEuropeanEngine is derived from its bases.

For this class, I provide member functions tranformOperator and 
transformBoundaryConditions.

I will also have the calculate() member function very much similar to 
the one in FDEuropeanEngine, with the exception that I will be calling 
rollback multiple times and calling both tranformOperator() and 
transformBoundaryConditions() just prior to that.

I am still writing the code now but I will like to hear if there are any 
comments, esp if this will work and is there a better way to do it for 
future flexibility.

Thanks!

Kuan