the LMM in QuantLib
Locked
 
|
Dear All,
I have started working on the LMM in QuantLib again. The main new addition is the Giles--Glasserman smoking adjoints method for delta and vega computation. I'd be interested in any feedback users have. eg 1) what's buggy? (if anything) 2) what features do you need that are missing? 3) do you find it hard to use? 4) if you aren't using it why not? For those who are finding it hard to follow the details, I will be doing a course in London in February on the QuantLib implementation of the LIBOR Market Model, we will do the whole thing from theory through to the C++ code. Book by Sep 30th for a special very early bird discount. http://www.moneyscience.com/Events_Noticeboard/article558 best Mark -- Quant Job Interview Questions and Answers is now out: www.markjoshi.com Assoc Prof Mark Joshi Centre for Actuarial Studies University of Melbourne My website is www.markjoshi.com ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ QuantLib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-users |
|
|
Mark,
It sounds interesting. I haven't used it yet. Currently, I am tied up with some projects. But, as soon as I shape things up, I would like to check it out. Do you have some material which may help to read the source code? Thank you, Javit
|
|
|
In reply to this post by Mark joshi
At the moment, the main thing I am doing to assist users is running the course in February which will discuss the C++ code:
http://www.moneyscience.com/Events_Noticeboard/article558 regards Mark ------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Mark, It sounds interesting. I haven't used it yet. Currently, I am tied up with some projects. But, as soon as I shape things up, I would like to check it out. Do you have some material which may help to read the source code? Thank you, Javit Mark joshi wrote: > > Dear All, > > I have started working on the LMM in QuantLib again. The main new > addition is the Giles--Glasserman smoking adjoints method for delta > and vega computation. > > I'd be interested in any feedback users have. > eg > 1) what's buggy? (if anything) > 2) what features do you need that are missing? > 3) do you find it hard to use? > 4) if you aren't using it why not? > > For those who are finding it hard to follow the details, I will be > doing a course in London in February on the QuantLib implementation of > the LIBOR Market Model, we will do the whole thing from theory through > to the C++ code. Book by Sep 30th for a special very early bird discount. > > http://www.moneyscience.com/Events_Noticeboard/article558 > > > best > > Mark > > > -- > Quant Job Interview Questions and Answers is now out: > www.markjoshi.com > > Assoc Prof Mark Joshi > University of Melbourne > My website is www.markjoshi.com > > ---------------------------------------------------------------------- > --- This SF.Net email is sponsored by the Moblin Your Move Developer's > challenge Build the coolest Linux based applications with Moblin SDK & > win great prizes Grand prize is a trip for two to an Open Source event > anywhere in the world > http://moblin-contest.org/redirect.php?banner_id=100&url=/ > _______________________________________________ > QuantLib-users mailing list > [hidden email] > https://lists.sourceforge.net/lists/listinfo/quantlib-users > > ----- Cavit (Javit) Hafizoglu mailto:[hidden email] mailto:[hidden email] -- View this message in context: http://www.nabble.com/the-LMM-in-QuantLib-tp19262300p19318695.html Sent from the quantlib-users mailing list archive at Nabble.com. -- Quant Job Interview Questions and Answers is now out: www.markjoshi.com Assoc Prof Mark Joshi Centre for Actuarial Studies University of Melbourne My website is www.markjoshi.com ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ QuantLib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-users |
|
|
In reply to this post by Mark joshi
Hi, Mark,
It seems that exerciseTimes are the same as rateTimes and evolutionTimes in the QuantLib code, although they should be different, especially the exerciseTimes, which usually are 10 - 30 business days before rateTimes. I'm referring to the classes such as MarketModelExerciseValue, MarketModelMultiProduct, LongstaffSchwartzExerciseStrategy. I also note that in collectNodeData, the node data are collected at evolutionTimes/rateTimes instead of exerciseTimes. But I think it should be exerciseTimes. Even we can choose evolutionTimes different from rateTimes and let exerciseTimes = evolutionTimes, the constructors of MarketModelExerciseValue and MarketModelMultiProduct have hard-coded rateTimes = evolutionTimes (except for the last rateTimes of course).
|
|
|
In reply to this post by Mark joshi
I am a bit mystified by this comment.
MarketModelExerciseValue and MarketModelMultiProduct don't have constructors since they are empty base classes. The code is written to run off EvolutionDescriptions where the evolution times and rate times are distinct concepts. There are certainly examples of MarketModelMultiProduct s with only one evolution time, for example, in the code. You can also specify which evolution times are exercise times. So you can merge rate times and exercise dates to get the set of evolution times for example. For early exercise, you need to be able to collect data at non-exercise times as well as exercise times in case coupons have path-dependence affected by non-exercise times. regards mark ------------------------------------------------------------------------------------------------ To: [hidden email] Date: Wed, 10 Sep 2008 21:10:36 -0700 (PDT) Subject: Re: [Quantlib-users] the LMM in QuantLib Hi, Mark, It seems that exerciseTimes are the same as rateTimes and evolutionTimes in the QuantLib code, although they should be different, especially the exerciseTimes, which usually are 10 - 30 business days before rateTimes. I'm referring to the classes such as MarketModelExerciseValue, MarketModelMultiProduct, LongstaffSchwartzExerciseStrategy. I also note that in collectNodeData, the node data are collected at evolutionTimes/rateTimes instead of exerciseTimes. But I think it should be exerciseTimes. Even we can choose evolutionTimes different from rateTimes and let exerciseTimes = evolutionTimes, the constructors of MarketModelExerciseValue and MarketModelMultiProduct have hard-coded rateTimes = evolutionTimes (except for the last rateTimes of course). Mark joshi wrote: > > Dear All, > > I have started working on the LMM in QuantLib again. The main new addition > is the Giles--Glasserman > smoking adjoints method for delta and vega computation. > > I'd be interested in any feedback users have. > eg > 1) what's buggy? (if anything) > 2) what features do you need that are missing? > 3) do you find it hard to use? > 4) if you aren't using it why not? > > For those who are finding it hard to follow the details, I will be doing a > course in London in February on the QuantLib implementation of the LIBOR > Market Model, we will do the whole thing from theory through to the C++ > code. Book by Sep 30th for a special very early bird discount. > > http://www.moneyscience.com/Events_Noticeboard/article558 > > > best > > Mark > > > -- > Quant Job Interview Questions and Answers is now out: www.markjoshi.com > > Assoc Prof Mark Joshi > Centre for Actuarial Studies > University of Melbourne > My website is www.markjoshi.com ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ QuantLib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-users |
|
|
I am a bit mystified by this comment.
MarketModelExerciseValue and MarketModelMultiProduct don't have constructors since they are empty base classes. The code is written to run off EvolutionDescriptions where the evolution times and rate times are distinct concepts. There are certainly examples of MarketModelMultiProduct s with only one evolution time, for example, in the code. You can also specify which evolution times are exercise times. So you can merge rate times and exercise dates to get the set of evolution times for example. For early exercise, you need to be able to collect data at non-exercise times as well as exercise times in case coupons have path-dependence affected by non-exercise times. regards mark ------------------------------------------------------------------------------------------------ To: [hidden email] Date: Wed, 10 Sep 2008 21:10:36 -0700 (PDT) Subject: Re: [Quantlib-users] the LMM in QuantLib Hi, Mark, It seems that exerciseTimes are the same as rateTimes and evolutionTimes in the QuantLib code, although they should be different, especially the exerciseTimes, which usually are 10 - 30 business days before rateTimes. I'm referring to the classes such as MarketModelExerciseValue, MarketModelMultiProduct, LongstaffSchwartzExerciseStrategy. I also note that in collectNodeData, the node data are collected at evolutionTimes/rateTimes instead of exerciseTimes. But I think it should be exerciseTimes. Even we can choose evolutionTimes different from rateTimes and let exerciseTimes = evolutionTimes, the constructors of MarketModelExerciseValue and MarketModelMultiProduct have hard-coded rateTimes = evolutionTimes (except for the last rateTimes of course). - Hide quoted text - Mark joshi wrote: > > Dear All, > > I have started working on the LMM in QuantLib again. The main new addition > is the Giles--Glasserman > smoking adjoints method for delta and vega computation. > > I'd be interested in any feedback users have. > eg > 1) what's buggy? (if anything) > 2) what features do you need that are missing? > 3) do you find it hard to use? > 4) if you aren't using it why not? > > For those who are finding it hard to follow the details, I will be doing a > course in London in February on the QuantLib implementation of the LIBOR > Market Model, we will do the whole thing from theory through to the C++ > code. Book by Sep 30th for a special very early bird discount. > > http://www.moneyscience.com/Events_Noticeboard/article558 > > > best > > Mark > > > -- > Quant Job Interview Questions and Answers is now out: www.markjoshi.com > > Assoc Prof Mark Joshi > Centre for Actuarial Studies > University of Melbourne > My website is www ------------------------------------------------------------------------- This SF.Net email is sponsored by the Moblin Your Move Developer's challenge Build the coolest Linux based applications with Moblin SDK & win great prizes Grand prize is a trip for two to an Open Source event anywhere in the world http://moblin-contest.org/redirect.php?banner_id=100&url=/ _______________________________________________ QuantLib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-users |
|
|
In reply to this post by Mark joshi-2
Sorry. I actually mean the derived classes of MarketModelExerciseValue and MarketModelMultiProduct.
Although i can choose different evolution times in derived class of MarketModel, the evolution defined in constructor such as MultiProductMultiStep, BermudanSwaptionExerciseValue, is the same as ratetimes. This can cause isBasisTime in collectNodeData to be all "false" if I use different evolution times for dataProvider such as SwapBasisSystem. isProductTime, isRebateTime, isControlTime, isExerciseTime are guranteed to be true since they are all from rate times. <quote author="Mark joshi-2"> I am a bit mystified by this comment. MarketModelExerciseValue and MarketModelMultiProduct don't have constructors since they are empty base classes. The code is written to run off EvolutionDescriptions where the evolution times and rate times are distinct concepts. There are certainly examples of MarketModelMultiProduct s with only one evolution time, for example, in the code. You can also specify which evolution times are exercise times. So you can merge rate times and exercise dates to get the set of evolution times for example. For early exercise, you need to be able to collect data at non-exercise times as well as exercise times in case coupons have path-dependence affected by non-exercise times. regards mark |
«
Return to quantlib-users
|
1 view|%1 views
| Free forum by Nabble | Edit this page |