SERVING THE QUANTITATIVE FINANCE COMMUNITY

  • 1
  • 4
  • 5
  • 6
  • 7
  • 8
  • 18
 
User avatar
Cuchulainn
Posts: 59926
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

The ultimate Monte Carlo framework

January 4th, 2012, 7:40 pm

Hi Alex,This looks very nice. I quickly ran some of the projects. Some seem to use libs which are not in the projects. How can I resolve this?It seems that we have 2 MC engines beginning to blossom. My idea is that we should share on . FD algorithms . Financial closed solutions. Extension to MC (e.g. antithetic,..)I have added the following to my MC101;. Predictor-Corrector (3 flavours + 1 new one based on midpoint rather than talking averages at n and n+1 !!). Richardson Euler. Milstein. Karhunen-Loeve with the above schemes (KL is like Fourier series and has smallest MSE). It essentially transforms Ito integral to a Lebesque-Stieltjes integral (I think...). Antithetic Euler. For KL (uses sin function) I use OpenMP for speedup (don't use SSE2 nor SIMD yet).What I want next is to design Jumps into the SDE.Problems being worked on: Plain, barrier, CIR, Asian. I will post MC102 soon and maybe you find useful stuff there.//Hope to get back to you when lib issue is resolved.
Last edited by Cuchulainn on January 3rd, 2012, 11:00 pm, edited 1 time in total.
 
User avatar
Cuchulainn
Posts: 59926
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

The ultimate Monte Carlo framework

January 4th, 2012, 8:11 pm

What do you think of the scheme in section 3.2?BTW instead of scheme 3.6 (ignoring jumps) I use a midpoint variant (on examples it gives better results). Does anyone have a corresponding PIDE solution for testing against?
Last edited by Cuchulainn on January 3rd, 2012, 11:00 pm, edited 1 time in total.
 
User avatar
Cuchulainn
Posts: 59926
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

The ultimate Monte Carlo framework

January 4th, 2012, 8:33 pm

QuoteOriginally posted by: outrunhi both:good progress indeed!you should really be starting to use the centralized code repository, it does versioning, gives people uniform and easy access.What do you need to start using that?Robert (the guy who cooks wonderful Nasi just had some today, after all it's Wednesday) will help this week with the sandbox. He can help me get up to speed with the Windows part.Do we need a separate Linux sandbox?I hope to be able to get some tuition as it is hard to focus on all the issues at one time.ThanksSome project mgt help would also be nice. My SDE stuff could be mapped to concepts I reckon.
Last edited by Cuchulainn on January 3rd, 2012, 11:00 pm, edited 1 time in total.
 
User avatar
rmax
Posts: 6080
Joined: December 8th, 2005, 9:31 am

The ultimate Monte Carlo framework

January 5th, 2012, 8:41 am

Depending on what you need more than happy to act as some form of Project Manager if Outrun wants to focus on more of the architecture/development side. Problem for me at the moment is time, but I can try and dedicate some time to this as I believe the project is an excellent idea - as long as we avoid the traps of QuantLib.
 
User avatar
rmax
Posts: 6080
Joined: December 8th, 2005, 9:31 am

The ultimate Monte Carlo framework

January 5th, 2012, 9:07 am

In essence yes as long as we get commitment/buy-in from people on decisions made etc. However there is a big elephant sitting in the room... Is this a Cathedral or a Bizarre? We keep skirting around the issue and I think this is something that is holding us in terms of peoples commitment.
 
User avatar
FinancialAlex
Posts: 81
Joined: April 11th, 2005, 10:34 pm

The ultimate Monte Carlo framework

January 5th, 2012, 10:33 am

QuoteI quickly ran some of the projects. Some seem to use libs which are not in the projects. How can I resolve this?Daniel, can you specify which libs you are encountering problems with? I have downloaded the zip and build the library using only the files in the zip, and was able to price the Asian call option.I have opened the solution found in the directory MonteCarloEngineConsole
 
User avatar
FinancialAlex
Posts: 81
Joined: April 11th, 2005, 10:34 pm

The ultimate Monte Carlo framework

January 5th, 2012, 10:37 am

QuoteDoes anyone have a corresponding PIDE solution for testing against? Two articles which have analytical solutions of PIDEs which were checked against (as far as I remember):Mayo A. Methods for the rapid solution of the pricing PIDEs in exponential and Merton models. J. Comp Appl. Math, Vol 222, pp. 128-143, 2008.Almendral A. and Oosterlee C. Accurate evaluation of European and American options under CGMY process. SIAM J Sci Comput., Vol 29, pp. 93-117, 2007.
 
User avatar
Cuchulainn
Posts: 59926
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

The ultimate Monte Carlo framework

January 5th, 2012, 11:38 am

QuoteOriginally posted by: FinancialAlexQuoteI quickly ran some of the projects. Some seem to use libs which are not in the projects. How can I resolve this?Daniel, can you specify which libs you are encountering problems with? I have downloaded the zip and build the library using only the files in the zip, and was able to price the Asian call option.I have opened the solution found in the directory MonteCarloEngineConsoleMy bad I now see that the others are class libs. BTW does the UML diagram you posted represent the current code?
 
User avatar
Cuchulainn
Posts: 59926
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

The ultimate Monte Carlo framework

January 8th, 2012, 3:08 pm

Boost and PRNG thread
 
User avatar
Cuchulainn
Posts: 59926
Joined: July 16th, 2004, 7:38 am
Location: Amsterdam
Contact:

The ultimate Monte Carlo framework

January 9th, 2012, 3:29 pm

MC102 = MC101 + a bunch of well-known, improved and new schemes. Architecture has not changed.Quizzies:1. Incorporate jumps.2. Price Asian, vanilla, barrier options in parallel, OpenMP, Boost How? (must be non-intrusive)3. Writing batch jobs to test schemes/serialisation. Do you need to extend/modify the architecture?4. Euler moment matching gives too low answers; bug or something else. dunno.Focus MC102 is numerical accuracy.
Attachments
MC102.zip
(30.95 KiB) Downloaded 12 times
Last edited by Cuchulainn on January 8th, 2012, 11:00 pm, edited 1 time in total.
ABOUT WILMOTT

PW by JB

Wilmott.com has been "Serving the Quantitative Finance Community" since 2001. Continued...


Twitter LinkedIn Instagram

JOBS BOARD

JOBS BOARD

Looking for a quant job, risk, algo trading,...? Browse jobs here...


GZIP: On