First SENSORFINT Training School in Chemometrics

Objectives

This training session on Chemometrics aims to deepen the trainee’s knowledge in spectral data analysis. Aspects of unsupervised analysis and multivariate linear and nonlinear calibration will be addressed. The pedagogical objective is to provide the trainees with the theoretical and practical bases to make them autonomous to manage their spectral data.
The training session proposes alternatively theoretical method principles and practical exercises using the ChemFlow software.

Dates

4 days – 28 hours
From 5 Sept -2022 2.00 pm to 9 Sept -2022 2.00 pm

Public:

PhD, Post doctorants
Engineers, Researchers

Pre-requisite:
Practice (or knowledge) in spectroscopy, to be used with spectral data
Practice on factorial methods, to be able to calibrate a PLS, to carry out a PCA, …

Program:

Introduction

Introduction to the tools (ChemFlow)

PCA Theory & application – Tom Fearn – 2 half days

Classical case (low inter correlation)
Spectral case (high inter correlation)
Outlier detection
Loading interpretation
Pretreatment selection

Unsupervised Challenge: choosing the best pretreatments, detecting outliers, structure, etc. without any model

Modelling & Validating – Marina Cocchi & Jean Michel Roger – 3 half days

PLS Algorithm without maths,
Notion of dimension, Figures of merit
Tuning of dimension, CV
Modelling workflow,
Preprocessings,
Case studies, including fails

Supervised Challenge: tune the best model, on the data preprocessed in session 2, test on unknown samples. Discussion

ANN based methods – Federico Marini & Eric Latrille – 1 half day

Introduction about ANN, self-organizing maps or GTM alternative method
Application on data set and software

Non linear Methods – Matthieu Lesnoff – 1 half day

Local methods
Application on data set and software

Final Challenge – 1 half day

Teachers:

Tom Fearn , from UCL – London’s Global University,UK
Tom Fearn is Professor of Applied Statistics in the Department of Statistical Science at UCL in London. His main research area is chemometrics, and in particular its applications to quantitative calibration and classification problems in NIR spectroscopy. He received the Tomas Hirschfeld award for contributions to NIRS in 2001 and is Immediate Past President of the International Council for Near Infrared Spectroscopy

Federico Marini, from La Sapienza Università di Roma, Italy

Prof. Marini received his MSc from Sapienza University of Rome in 2000 where he also completed his PhD in 2004. In 2006, he was awarded the Young Researcher Prize from Italian Chemical Society and in 2012 he won the Chemometrics and Intelligent Laboratory Systems Award “for his achievements in chemometrics”. His research activity is focused on all aspects of chemometrics, ranging from the application of existing methods to real world problems in different fields to the design and development of novel algorithms. He is member of the Editorial boards of Chemometrics and Intelligent Laboratory Systems, Analytica Chimica Acta, J. of Chemometrics, J. of NIR Spectroscopy, J. of Spectral Imaging and he serves as Associate Editor for Chemometrics in Wiley’s Encyclopedia of Analytical Chemistry. He is the past-coordinator of the Chemometric group of the Italian Chemical Society, and currently the vice-president of Italian NIR society and the coordinator of the Chemometric study group of EUCheMS

Marina Cocchi from University of Modena and Reggio Emilia, Italy
Marina Cocchi is at present Associate Professor of Analytical Chemistry at University of Modena and Reggio Emilia. She got the PhD degree in Chemical Sciences in 1992 at the same University. Author of more than 100 publications on International Journals and books (H-ind 27, 2246 cit., Scopus). Main research topics: Multivariate, Multi-way and Multiset methods; Data Fusion; Feature selection; MSPC; Food Authenticity; Chemical fingerprinting by spectroscopy (MIR, NIR, NMR) and chromatography; Multivariate calibration; Experimental Design. She is member of editorial Board Chemometrics and Intelligent Laboratory Systems, and of Advisory Board of Comprehensive Chemometrics, Wiley, 2nd Ed.

Jean Michel Roger from INRAE, France
Jean-Michel Roger focused is research on the robustness problem. In 2003, simultaneously with Prof. Tom Fearn, he proposed a new generic method called External Parameter Orthogonalization (EPO) which permits to eliminate specific spectral effects from the NIRS calibration and thus to solve certain problems of robustness.

He is also involved in the design of portable or online spectrometers using filters technology. In this aim, he developed a method (CovSel) that allows to select wavelengths with regard to multiple responses. He has created ChemHouse, a chemometrics research group, forge of the ChemFlow software (see https://chemproject.org).

Matthieu Lesnoff
Matthieu Lesnoff is biometrician at the Selmet (Mediterranean and tropical livestock farming systems) research unit in CIRAD. In chemometrics, he specialized on non-linear prediction methods, in particular on locally weighted PLSR approaches. He is actually in charge of developing a web platform for automated predictions of the composition of tropical forage compositions in Cirad. He is member of the ChemHouse research group in Montpellier (https://chemproject.org) and has developed several chemometrics free packages in R and Julia languages (rchemo and Jchemo: https://github.com/mlesnoff)

Software practice supervisors

	Jean Claude Boulet	UMR SPO, INRAE Montpellier
	Gilles Chaix	UMR AGAP, CIRAD Montpellier
	Martin Ecarnot	UMR AGAP, INRAE Montpellier
	Eric Latrille	UR LBE, INRAE Montpellier
	Virginie Rossard	UR LBE, INRAE Montpellier