Bugs item #1121342, was opened at 2005-02-12 15:37
Message generated for change (Settings changed) made by lballabio
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=112740&aid=1121342&group_id=12740Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: None
Group: None
>Status: Closed
>Resolution: Works For Me
Priority: 5
Private: No
Submitted By: Nobody/Anonymous (nobody)
>Assigned to: Ferdinando Ametrano (nando)
Summary: Implementation of Monotone Spline
Initial Comment:
My email address is
[hidden email]
The implementation of the Hyman filter for the
monotonising of the cubic spline is wrong. The Hyman
constraints are corrrectly specified but are applied to the
second derivative of the function (y'') rather than to y'.
In order to implement correctly you need to
i) solve for the y'' values at the node in the usual way
ii) compute the first derivative values y' at each node
iii) generate the appropriate lagrange cubic polynomials
through the node points with the correct y' value at the
ends.
The incorrect approach does not guarantee to produce
monotone splines.
----------------------------------------------------------------------
Comment By: Ferdinando Ametrano (nando)
Date: 2005-02-14 11:20
Message:
Logged In: YES
user_id=34616
why do you say that the contraints are "applied to the
second derivative of the function (y'') rather than to y'." ?
I wrote the code some time ago and I might be rusty, but it
looks to me that the Hyman filter is correctly applied to
the first derivative.
Beside the test suite is able to correctly reproduce Hyman
numerical examples.
Do you have a numerical examples to show a possible
monotonicity failure of the current implementation?
thank you
ciao -- Nando
----------------------------------------------------------------------
You can respond by visiting:
https://sourceforge.net/tracker/?func=detail&atid=112740&aid=1121342&group_id=12740-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/_______________________________________________
QuantLib-dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-dev
Locked
