STAT IR9

School-StatIR9-2027-Bandeau-EN.png
May 26-28 2027 • Synchrotron SOLEIL

The SMIS beamline at the SOLEIL synchrotron is pleased to announce the 9th edition of its training workshop, STAT-IR 9, focused on multivariate and machine learning analysis of infrared microspectroscopy data. The three-day course will take place from Wednesday, May 26th to Friday, June 28th, 2027, at the SOLEIL synchrotron and will be conducted in English.

The workshop aims to provide participants with hands-on experience on the use of Multivariate Analysis (MVA) techniques such as PCA, and Machine Learning (ML) methods including both unsupervised and supervised clustering approaches for infrared microspectroscopy data analysis. The course will also cover spectral preprocessing techniques used in preparation for MVA and ML. Additionally, the course will cover the use of these techniques in analyzing hyperspectral images and maps. The training program will emphasize biomedical datasets, though the methodologies are equally applicable to data from cultural heritage and polymer sciences. These techniques can also be extended to data obtained from Raman microspectroscopy and X-ray fluorescence.

Infrared microspectroscopy and imaging provide essential insights into the chemical composition, spatial organization, and molecular conformation of complex materials with spatial resolutions ranging from microns to submicron. Efficient data analysis is key to unlocking the rich information embedded in infrared spectra, especially when dealing with large datasets comprising tens to hundreds of thousands of spectra.  This requires the use of automated techniques to effectively extract meaningful information.

One of the most effective approaches to handling large datasets is statistical analysis which can reveal spectral variability, identify correlations, classify data objectively, train identification algorithms, integrate data from multiple techniques, and establish predictive models for quantification. 

The success of these methods hinges on the application of appropriate spectral preprocessing to eliminate artifacts and confusing variations. Proper data analysis strategies must be employed to validate statistical models and guarantee that they are robust and generalizable. Mastery of these specialized software tools and techniques requires dedicated training in signal processing that will be taught during the workshop.

The SMIS beamline at the SOLEIL Synchrotron, located near Paris, has over 15 years of expertise in the measurement and analysis of microspectroscopy data. This course will be led by:

    Ferenc Borondics, expert in physical chemistry,

    Marko Toplak, machine learning specialist and one of the developers of Quasar software,

    Christophe Sandt, biospectroscopy expert.

The course will utilize the open-source software Quasar, and participants will have the opportunity for hands-on practice with their own data.

The workshop will take place on-site at the SOLEIL synchrotron and is open to all past and future users of the SMIS beamline. The registration fee is 120 €, which includes on-site meals.