Product Exponential Linear Fit Function

[Owen Arnold]

Following #5626 we need functions specific to detector efficiency corrections. These need to be named functions with named parameters so that they can be configured in the IDF xml.

[Owen Arnold]

refs #5633. New parameter function produced

Changeset: 0b7a4be579ef2ea24d4d3c10255c2d25e9f43308

[Owen Arnold]

refs #5633. Working and tested param function

Changeset: e6ab7071fe373fd11bf4a0259e740d5fe546ef52

[Owen Arnold]

refs #5633. Fix GCC errors and warnings

Changeset: e7d689242d1638f92a6f96f1cb094794b201f52d

[Owen Arnold]

refs #5633. Fix GCC errors

Changeset: dbb8db89633f7d88d226afcec9f5bf804c68234d

[Owen Arnold]

refs #5633. Fix test precision issue on mac

Changeset: fe70e69e5718be81cb84cd630fca47b10f788258

[Owen Arnold]

refs #5633. Characterise derivs using mocking

Changeset: 3d8e525bcb9637257e3efb8ce02ede1947193d98

[Owen Arnold]

Create a wiki page to describe the algorithm @ http://www.mantidproject.org/ProductLinearExp

[Owen Arnold]

refs #5633. Calculate Jacobian.

Changeset: 60d8b749ecebe1a1bd802b5ecdb398bfdc79fac0

[Owen Arnold]

refs #5633 #5634. Documentation.

Changeset: 3b1821e1920f87f9756d2accbf92cb8ba758c52f

[Owen Arnold]

Tester: The following python code (along with the attached demo.xml file) can be used to verify that this fit function is working, and is compatible with the NormaliseByDetector algorithm, for which it has been written. Running the following script should not cause any exception to be raised.

import math 

inWS = Load(Filename='POLREF00004699.nxs') # in AutoTest dir
inWS = ConvertSpectrumAxis(InputWorkspace=inWS,Target='signed_theta')
inWS = inWS[0]
inWS = ConvertUnits(InputWorkspace=inWS,Target='Wavelength',EMode='Indirect')

# Overwrite the
for i in range(0, inWS.getNumberHistograms()):
        Y = inWS.dataY(i)
        for j in range(0, len(Y)):
                Y[j] = 1

# Parameter file sets a global ProductLinearExp fit function for the whole instrument.
LoadParameterFile(Workspace=inWS,Filename=r'demo.xml')
normalised = NormaliseByDetector(InputWorkspace=inWS)

# It's enough just to use just the first histogram and compare the expected Y values with the Actual ones to determine if the ProductLinearExp fit function is being used.
for k in range(0, 1):
        Xout = normalised.dataX(i)
        Yout = normalised.dataY(i)
        Yin = inWS.dataY(i)
        for l in range(0, len(Xout)-1):
                x = (Xout[l]  + Xout[l + 1])  / 2
                yExpected= 1/math.exp(-x/1) # because we set A1 = 0, A0 = 1. Height=1 and Lifetime = 1 in the demo.xml file.
                yActual = Yout[l]
                if((math.fabs(yExpected) - math.fabs(yActual)) > 0.0001):
                        print math.fabs(yExpected) , math.fabs(yActual)
                        raise RuntimeError("Error, NormaliseByDetector is not using ProductLinearExp fit function properly.")

[Owen Arnold]

refs #5633 #5634. Test via FunctionFactory

Quite a simple but important test, because these functions would not work via NormaliseByDetector unless they were factory registered. Also provides checking that they are types of IFunction, since the test downcasts the boost shared_ptr.

Changeset: 76e97bb7f0c883f83d6a97f19d0ac2b1f91ab3fe

[Michael Reuter]

This seems to be working fine.

[Stuart Campbell]

This ticket has been transferred to github issue ​6479