Affiliations: ILL, Grenoble, France | Technical University of Denmark, Kongens Lyngby, Denmark
Note: [] Corresponding author: E. Farhi, ILL, BP 156, 38042 Grenoble Cedex 9, France. Tel.: +33 476 207 135; E-mail: farhi@ill.fr
Abstract: We present a new tool, iFit, which uses a single object class to hold any data set, and provides an extensive list of methods to import and export data, view, manipulate, apply mathematical operators, optimize problems and fit models to the data sets. Currently implemented using Matlab®, the toolbox is lightweight and comes with an extensive documentation based on tutorials with ready-to-run examples for each operator. Provided with the package is a set of optimization algorithms, which we have benchmarked in order to recommend the ones that provide the best success rate for both continuous and noisy problems. These optimizers can then be used to fit models onto data objects, and optimize McStas instrument simulations. As an application, we propose a methodology to analyse neutron scattering measurements in a pure Monte Carlo optimization procedure using McStas and iFit. As opposed to the conventional data reduction and analysis procedures, this new methodology is able to intrinsically account for most of the experimental effects, and results in the sample only model, de-convolved from the instrument.
Keywords: Numerical analysis, optimization, Monte Carlo, neutron scattering
