Ticket #983 (closed: fixed)

Opened 11 years ago

Last modified 5 years ago

Add peakshape functions which satisfy IPeakFunction interface

Reported by: Anders Markvardsen Owned by: Anders Markvardsen
Priority: major Milestone: Iteration 21
Component: Keywords:
Cc: nick.draper@… Blocked By:
Blocking: Tester:

Description

Migrate curvefitting peakshape functions to new IPeakFunction interface.

Change History

comment:1 Changed 11 years ago by Anders Markvardsen

(In [3305]) Add Gaussian to new IPeakFunction interface. Refs #983

comment:2 Changed 11 years ago by Anders Markvardsen

(In [3309]) Add test to Gaussian. Refs #983

comment:3 Changed 11 years ago by Anders Markvardsen

(In [3317]) Add quadratic IFunction and preliminary test. Refs #983

comment:4 Changed 11 years ago by Anders Markvardsen

(In [3318]) Add quadratic IFunction with lower case hence the reason for not building on Linux. Will delete then add them again. Refs #983

comment:5 Changed 11 years ago by Anders Markvardsen

(In [3319]) Add quadratic IFunction with upper case file names. Refs #983

comment:6 Changed 11 years ago by Anders Markvardsen

(In [3322]) The QuadraticTest studied in more detail. The Quadratic IFunction is linear in the parameters and it is found that the GSL Levenberg-Marquardt algorithm designed for nonlinear miminisation is struggling to handle the Quadratic fitting function. For example when an x-range interval bigger than 1000 is used. A working test with an interval of 100 is here submitted. Refs #983

comment:7 Changed 11 years ago by Anders Markvardsen

(In [3324]) Add LinearBackground satisfying IFunction. Note that this function suffers for now from the same limitation as the Quadratic IFunction. Refs #983.

comment:8 Changed 11 years ago by Anders Markvardsen

(In [3335]) Solved problem with linear fitting misbehaving in LinearBackground and Quadratic I believe. It turned out that by coincidence I was fitting on an interval of HRPD data for which a few of the data points have their error value assigned to zero!!!!!! Hence no big surprise that the fitting went bad for this case. Have changed Fit1D such that if a data point with zero error is encountered then an error of 1 is assigned to such a data point! Refs #983.

comment:9 Changed 11 years ago by Anders Markvardsen

(In [3340]) Add Lorentzian peak shape function. Refs #983

comment:10 Changed 11 years ago by Anders Markvardsen

(In [3377]) Add Ikeda Carpenter satisfying IPeakFunction. Refs #983.

comment:11 Changed 11 years ago by Anders Markvardsen

(In [3384]) Added test for Lorentzian. Added more sensible implementations of setWidth() etc methods for Gaussian and IkedaCarpenter. Refs #983.

comment:12 Changed 11 years ago by Anders Markvardsen

(In [3401]) Added IPeakFunction version of BackToBackExponential and updated rest of old IPeakFunctions. Refs #983.

comment:13 Changed 11 years ago by Anders Markvardsen

(In [3404]) Spotted problem with the way zero error values were handled and updated fit function tests accordingly. Refs #983

comment:14 Changed 11 years ago by Anders Markvardsen

(In [3406]) Fix broken LorentzianTest added in last commit to this ticket. Refs #983.

comment:15 Changed 11 years ago by Anders Markvardsen

(In [3407]) Add tests for BackToBackExponential. Refs #983.

comment:16 Changed 11 years ago by Martyn Gigg

(In [3409]) Removed debug code that was causing a few tests to report that they were failing when they weren't. Re #983

comment:17 Changed 11 years ago by Anders Markvardsen

  • Status changed from new to accepted

comment:18 Changed 11 years ago by Anders Markvardsen

  • Status changed from accepted to testing
  • Resolution set to fixed

Transferred all previous defined peakshape function to IPeakFunction

comment:19 Changed 11 years ago by Martyn Gigg

  • Status changed from testing to closed

comment:20 Changed 5 years ago by Stuart Campbell

This ticket has been transferred to github issue ​1831

Note: See TracTickets for help on using tickets.