I've uploaded a new fdeuropean.cpp and fdeuropean.hpp to the Wiki at
http://wiki.quantlib.org/twiki/bin/view/Quantlib/FiniteDifferencePricingEnginesThe new code is build on top of a new class called fdvanillaengine.cpp
that copies a lot of fdbsmoption.cpp. One the the big differences is
that the fd engines only calculate the greeks which are "easy" to get
from finite differencing (delta and gamma).
The next step is to calculate american and bermudan options. I'd like
to work on the FiniteDifference engine to create some new classes that
can handle coupled PDE's (essentially a list of operators which iterate
over a list of arrays), and I'll try to use that framework to calculate
control variates.
Question about greeks. Each calculation method has greeks that are
"easy" to calculate. For example if you wanted a price curve or delta
curve, it would be much easier to calculate this via PDE's than
iterating over the binomial or monte-carlo. How does quantlib put this
"knowledge" into the pricing engines? And is there a framework for
calculating greeks that aren't easily calculated with a single engine.
(For example, the new fd pricing engines don't calculate vega or rho
since this isn't "easy" using PDE's).
Locked
