I just tried fitting your dataset in SasView, Pas. I don't have a back-to-back exponential model (one could be written) so used a Lorentzian peak model instead. When I used our L-M optimiser I had to give it a starting peak position 'in the ball park' before it would latch on. No surprise there. Then I cleared that fit and tried again with our DREAM optimiser. https://www.sasview.org/docs/user/qtgui/Perspectives/Fitting/optimizer.html On the first attempt I was trying to optimise scale, background and peak position at the same time. This failed to find a good solution, but the output showed me that scale and background were heavily correlated. So I cleared that fit and tried again with DREAM but this time I fixed background at 0.0. I then deliberately set the starting values of scale and peak position to 1.0; ie, way off the expected solution. DREAM got the fit in seconds. I then refined the background and width. The SasView optimisers are from a NIST package called bumps. It's available on pypi. Steve From: Pascal Manuel - STFC UKRI <pascal.manuel(a)stfc.ac.uk> Sent: 23 November 2023 17:13 To: Hampson, Thomas (STFC,RAL,ISIS) <thomas.hampson(a)stfc.ac.uk>uk>; mpb(a)mantidproject.org Subject: [Mpb] Re: Mantid fitting engine replacement Dear Tom, Here is an example of data, python script generated from mantid fitting (sent as txt as our mailing system does not like sending .py file) and a plot that shows how bad the fit is, although mantid believe this is a perfectly acceptable fit. If you get the starting parameters near spot on, it will eventually refine. This is trying to fit a back to back expo, typical of time of flight data on a diffractometer (but I have seen similar things with simple gaussian!) Other programs (also using LM methods) do a good job. in particular, the python fitting routines seem to do the job. Hope this helps, P Dr Pascal Manuel ISIS Pulsed Neutron and Muon Facility STFC Rutherford Appleton Laboratory Chilton Oxfordshire OX11 0QX United Kingdom Office: + 44 (0)1235 446290 Beamline : +44 (0)1235 567080 Mobile: +44(0) 7833 965011 (also for WhatsApp) ZoomID : https://ukri.zoom.us/j/5735149151 Skype : pascal.manuel73 ________________________________ From: Hampson, Thomas (STFC,RAL,ISIS) <thomas.hampson(a)stfc.ac.uk> Sent: Wednesday, November 22, 2023 5:32 PM To: ISIS Instrument Scientist <ISISInstsci(a)stfc.ac.uk> Cc: Foxley, Sarah (STFC,RAL,ISIS) <sarah.foxley(a)stfc.ac.uk>uk>; Farooq, Rabiya (STFC,RAL,ISIS) <Rabiya.Farooq(a)stfc.ac.uk>uk>; Haigh, Jonathan (STFC,RAL,ISIS) <Jonathan.Haigh(a)stfc.ac.uk> Subject: RE: Mantid fitting engine replacement Hi All, Many thanks to those of you who have replied so far with examples that demonstrate the limitations of our current fitting framework. I will respond to you individually shortly to make sure I have enough information to add them to the fitbenchmarking dataset. If anyone else has any examples of problems they've had with fitting in Mantid then I would be grateful for your response. More data will help us make a better informed decision about the potential replacement of the fitting engine. Regards, Tom Hampson From: Hampson, Thomas (STFC,RAL,ISIS) Sent: 08 November 2023 12:38 To: ISIS Instrument Scientist <ISISInstsci(a)stfc.ac.uk> Cc: Foxley, Sarah (STFC,RAL,ISIS) <sarah.foxley(a)stfc.ac.uk>uk>; Farooq, Rabiya (STFC,RAL,ISIS) <Rabiya.Farooq(a)stfc.ac.uk>uk>; Haigh, Jonathan (STFC,RAL,ISIS) <jonathan.haigh(a)stfc.ac.uk> Subject: Mantid fitting engine replacement Dear ISIS Instrument Scientists, We are in the process of evaluating potential replacements for the Mantid fitting engine and need your input. We are seeking real examples that will help us quantify the limitations of the current engine. Specifically, we would appreciate your feedback on any issues you've encountered with the Mantid fitting engine, such as: - Sensitivity to initial parameter values - Slow performance - Getting stuck in local minima To assist us in this evaluation, please provide examples with the following information: - Fit function used - Data used in the fit - Fit range - Initial parameter values - Any additional information you believe is relevant The data you provide will be used with the FitBenchmarking tool, maintained by RSE in ISIS. This will help us determine whether there is a more suitable replacement. We also welcome any general feedback on the fitting in Mantid and the fitting interface. The deadline for your input is the end of November. Your contributions will greatly assist us in making informed decisions about the future of the Mantid fitting engine. Thank you for your assistance. Sincerely, Tom Hampson (on behalf of the Mantid team)