Objectives
Interpretability is a key issue to develop insightful statistical and machine learning approaches in business and industry.
The workshop to be held in July 2021 in Naples, is based on 3 pillars:
Explore the connections between machine learning tools, sensitivity analysis and rule-based systems.
Exploit the contribution of generalized additive models for the development and visualization of interpretable statistical models.
Analyze and propose monitoring tools for Additive Manufacturing (AM) systems.
A half-day is dedicated to each topic, and will offer deep methodological insights together with real-world industrial motivations. The academic state of the art together with the industrial motivations and motivations would be complement by related software resources or specific applications or extensions. In addition, each half-day will end with a round-table providing a closing discussion challenging the different views of interpretability while addressing general issues like:
Co-chairs
Antonio Lepore, University of Naples Federico II, Italy
Biagio Palumbo, University of Naples Federico II, Italy
Jean-Michel Poggi, University of Paris-Saclay, Orsay, France
Authors/speakers list (in alphabetic order)
Clément Benard (Safran, France)
Fabio Centofanti (University of Naples Federico II, Italy)
Bianca Maria Colosimo (Politecnico di Milano, Italy)
Sébastien Da Veiga (Safran Tech, France)
Matteo Fasiolo (University of Bristol, UK)
Yannig Goude (EDF Lab, France)
Bertrand Iooss (EDF R&D, France)
Ron Kenett (KPA Group and Samuel Neaman Institute, Technion, Israel)
Andrea Palumbo (Avio Aero, Cameri, Novara, Italy)
Piercesare Secchi (Politecnico di Milano, Italy)
Claudia Schipani (Avio Aero, Cameri, Novara, Italy)
Erwan Scornet (Ecole Polytechnique, France)
Simon Wood (University of Edinburgh, UK)
Format
The format is online (without fees) due to the covid-19 pandemic situation.
Publication A book edited by Springer is already planned after the workshop.
Webinar It will be possible to access the talks at another time after the workshop through the ENBIS media center.
Local Organizing Committee chairs: Antonio Lepore and Biagio Palumbo.
Contact
For further details, please contact antonio.lepore@unina.it or biagio.palumbo@unina.it or
Jean-Michel.Poggi@math.u-psud.fr
Program
Each half-day is organized according to the same schedule:
90’ Talk1+Talk2 including 15’ discussion
45’ Coffee break
30’ Talk3 including 5’ discussion
45’ Round table
Monday 12/07/2021
Pillar 1: Interpretability via random forests
Explore the connections between machine learning, rule-based systems and sensitivity analysis.
09:00 - 10:30 Sébastien Da Veiga, Erwan Scornet: “Two approaches of interpretability: simple ML models vs black-box explainability”
11:15 - 11:45 Clément Bénard: “Black-box explainability: variable importance in random forests”
11:45 - 12:30 Round table chaired by Bertrand Iooss
Pillar 2: Interpretability via additive models
Exploit the contribution of generalized additive models for the development and visualization of interpretable statistical models.
14:00 - 15:30 Simon Wood: "Generalised additive models: interpretability via additivity"
Yannig Goude: "Interpretability of machine learning approaches for electricity demand forecasting"
16:15 - 16:45 Matteo Fasiolo: "Visual tools for additive model development and checking"
16:45 - 17:30 Round table chaired by Piercesare Secchi
Tuesday 13/07/2021
Pillar 3: Interpretability of big data in Industry 4.0
Analyze and propose monitoring tools for Additive Manufacturing (AM) systems.
09:00 - 10:30 Bianca Maria Colosimo: “In-situ data mining in Industry 4.0 - opportunities and challenges”
Andrea Palumbo, Claudia Schipani: “Digital for Additive Manufacturing: the case study of Avio Aero”
11:15 - 11:45 Fabio Centofanti: “Robust functional ANOVA with application to Additive Manufacturing”
11:45 - 12:30 Round table chaired by Ron Kenett