Finite Difference Engine for Option under CoxIngersollRoss Short Rate Process #824
Conversation
| void FdmCIRMixedPart::setTime(Time t1, Time t2) { | ||
| const Real v = std::sqrt(sigma1_->blackForwardVariance(t1, t2, strike_)/(t2-t1)); | ||
| const NinePointLinearOp op(dyMap_.mult(Array(mesher_->layout()->size(), v)));//TODO fix this | ||
| mapT_.reset(op); |
There was a problem hiding this comment.
Since you're about to throw op away at the end of this method, you can write mapT_.swap(op). This removes the need for the reset method.
There was a problem hiding this comment.
Thanks. I'm now getting a n error
unknown location(0): fatal error: in "QuantLib test suite/Finite Difference CIR tests/QuantLib__detail__quantlib_test_case(&FdCIRTest__testFdmCIRConvergence)": std::runtime_error: Boost assertion failed: buckets_
How do I fix this? I don't get the same error when I tried running locally.
|
I can't reproduce it either, but according to the Travis builds it's a memory access violation. Maybe an array access out of bounds somewhere? (On our machines it might just read uninitialized memory without complaining.) |
Initial Commit Cleanup code Add delta,gamma,theta calculations Fix build Remove unused variables Remove unused variables Reorder variables fix build fix build Fix mispell Fix stuff test fix
|
Not sure if it can help, but configuring with |
I tried doing a ./configure --enable-extra-safety-checks, make, make install and then run the test locally using CMake build but it still doesn't throw the error. Is this the right way to do it to get the CMake build to pick up the flag? Or should i set that option somewhere else for CMake? |
|
Not sure if CMake can pick it up. Passing an additional define ( |
|
I should have found the problem. The engine stored the underlying processes by reference, and sometimes they ended up dangling. I've pushed a fix. |
Thanks! |
The short rate for a CoxIngersollRoss process can be described by the following SDE
By following the methodology described here http://scriptiesonline.uba.uva.nl/document/656101, using Ito's Lemma and riskless portfolio, the PDE for an option under CIR short rate can be derived as
By doing a change of variables for X=lnS, the PDE becomes
A test has been added to show that the value converges between multiple evolver schemes.
Also, as the sigma for the underlying black scholes process can only be determined during setTime(t1, t2), I needed a way to update the coefficients in the ninepointlinearop(equivalent to the apxb in triplebandlinearop) so i created a reset method for that but I'm open to better ways of doing that instead.