[Quantlib-dev] a few questions on trees and diffusion process

[Ferdinando A. <na...@qu...>]

Hi Sad

you might have noticed that I'm using quite a few of your classes lately: 
Exercise, DiffusionProcess, TimeGrig, etc.
I'm especially interested in the medium term to "merge" into the Lattice 
framework the finite difference approach along with the tree based approach.

I've also added a few binomial trees that I coded for a course. A few 
things are not completely clear to me, and are stopping me from further 
work with the Lattice framework.
1) what is the addTimes method in charge of?
2) in general I see the rollback procedure too deeply encapsulated, not 
allowing easy intermediate rollback. Am I wrong? do you have a suggestion? 
E.g. I would like to perform a so-called black-scholes adjustment to all 
binomial trees, and that is an adjustment you perform only at the first 
rollback (T-1*dt), then you roll all the way back as usual. I'm quite 
confused where I could perform such adjustment

On a different topic: I would like to have a date interface for 
DiffusionProcess, along with the existing time interface. This is quite 
natural and easy when you take a look at the current BlackScholesProcess 
that accepts term structures as input. It is not that easy  for the 
OrnsteinUhlenbeckProcess, at least for me considering the way it is used in 
the model dynamics.
I performed a cursory look at the code, but where 
OrnsteinUhlenbeckProcesses are instantiated there are no Dates or 
TermStructure around providing the data I would need, basically the origin 
date and the dayCounter.
BTW to make OrnsteinUhlenbeckProcess accept time dependent parameters would 
allow to have interest rate models with non-flat vol and mean reversion