[Quantlib-dev] problems with localVolImpl

[Paulius J. <pau...@gm...>]

Dear all,

When getting a local volatility from implied volatility via
localVolImpl() (for example using a Monte Carlo simulation),
from time to time one is given a negative local volatility (which generates
an exception - as expected).
The problem often arises  from small numerical inaccuracies which get
augmented when numerically calculating the second derivative.

Has anyone  already encountered and resolved the problem?
If not, I'd like to contribute a fix to the library. This would mean
changing the interface s.t. the
LocalVolTermStructure::localVolImpl method returns a discretized vol - not
an instantaneous vol.
This would be consistent with the library structure - as it is the Euler
discretization method which
calls the local vol method.

Does anyone have any other suggestions?

Sincerely,
Paulius