Posted by
Zakoian2000 on
Jul 09, 2003; 5:13pm
URL: http://quantlib.414.s1.nabble.com/Re-bermudan-swaption-simplex-tp2586p2589.html
Is the simplex optimization still valid for more than 2 dimensions? It does
not work with CIR+ or G2++ and I don't know yet where the problem is.... Has
anybody made it working with the CIR + Jamshidian Swaption?
Regards,
Zakoian
-----Original Message-----
From:
[hidden email]
[mailto:
[hidden email]]On Behalf Of Giancarlo
Pfeifer
Sent: Friday, June 27, 2003 4:09 PM
To:
[hidden email]
Subject: [Quantlib-users] Re: bermudan swaption, simplex
> At 11:15 PM 2/4/03 +0100, roland.lichters@we... wrote:
> >I am trying to run the bermudan swaption example, but the Hull White
> >numerical calibration hangs.
>
> Roland,
> are you sure it actually hangs? On my box, the analytic
> calibration is almost instantaneous, while the numerical calibration takes
> over 10 minutes to complete---which might as well give the impression of >
hanging (and suggests choosing the analytic calibration for actual work :)
> Could you try and let it run during a lunch break or something?
>
> Oh, and thanks for the kind words...
>
> Later,
> Luigi
Thanks for creating QuantLib. I started use it a couple of weeks ago, and I
am impressed by the coding quality (expecially the consistency in which
design patterns are used).
I am now experimenting with Bermudan Swaptions and I have four questions...
1. I am trying to calibrate Hull-White for Bermudan swaptions, and the time
needed for calibration varies greatly depending on the volatility surface I
am inserting. For some surfaces, the analytic formula only takes a few
seconds, while with other surfaces the calibration does never seem to
terminate.
I am wondering if this could be due to the value of the parameter "lambda"
passed to the Simplex object.
In the demo, lambda is set to 0.25, and this might be an optimal value for
that particular surface, but maybe it is not optimal for other volatility
surfaces. Could this be the cause of the long time needed for some
calibrations? If so, is there a way to determine a good "lambda"?
Or should I expect the time used by the calibration to be more or less the
same, regardless of the volatility surface used? (in which case I am clearly
doing something wrong in my code?)
2. I am also wondering if the order in which the volatilities are inserted
matters. The demo has the following line:
swaption.push_back(
Handle<CalibrationHelpers>(
new SwaptionHelper(swaptionMaturities[j],
Period(swaptionLengths[i]), Years),
RelinkableHandle<MarketElement>(Vol),
indexSixMonths,
rhTermStructure)
that is called for each instrument to calibrate. In a old email, Francesco
Perissin (2002-09-06) suggested:
"pay attention to the vector of input swaption volatilities, sort them
correctly"
Is this correct? Why is sorting necessary (since maturity and tenor is
passed with vol)? And which kind of sorting? Shall the instruments be
inserted Increasing in swap tenor?
3. How does the SwaptionHelper class works? Looking at the implementation,
it seems to me, that it generates the swap that corresponds to the ATM of
the swaption, but I notice that the frequencies for the floating and fixed
legs are both set to the frequency of the index, and the day counting
convention is the one for the index too.
Does it mean that, for example for an EURO swaption, the swaption volatility
surface I am using should be computed for swaptions semi-annual on both the
fixed and the floating leg? (if I am not wrong, people tend to quote vols
for annual bond vs 6m....). This looks a bit strange to me... Am I wrong?
4. Even when I manage to calibrate the surface with Hull&White, the
volatilities can differ for quite big amounts. I guess that what I am
looking for is an implementation of Hull and White with alpha that depends
on time... Is anyone working on the implementation of the Hull&White with
non constant alpha? If not, have you any suggestions in case I wanted to try
to create a new class in Quantlib to implement it? (I don't know if I will
have enough time to do it immediately, but I am thinking about it...).
Thanks a lot for your patience reading this and for all the effort you are
putting on Quantlib.
gc
****************************************************************************
***
For the latest news about ICAP go to www.icap.com.
The information in this Internet e-mail is confidential and may be legally
privileged. It is intended solely for the addressee. Access to this
Internet e-mail by anyone else is unauthorised and any disclosure, copying,
distribution or any action taken or omitted to be taken in reliance on it,
is prohibited and may be unlawful.
When addressed to our clients any opinions or advice contained in this
Internet e-mail are subject to the terms and conditions expressed in any
applicable documentation or market practices governing the relationship
between ICAP plc and its clients.
We have taken precautions to minimise the risk of transmitting software
viruses, but we advise you to carry out your own virus checks on any
attachments to this message. We cannot accept liability for any loss or
damage caused by software viruses.
Any views expressed in this message are those of the individual sender
except where they are stated to be the views of ICAP plc.
****************************************************************************
***
-------------------------------------------------------
This SF.Net email is sponsored by: INetU
Attention Web Developers & Consultants: Become An INetU Hosting Partner.
Refer Dedicated Servers. We Manage Them. You Get 10% Monthly Commission!
INetU Dedicated Managed Hosting
http://www.inetu.net/partner/index.php_______________________________________________
Quantlib-users mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-users