Abstract: 

I will briefly recall why explaining the predictions of a complex machine learning model (for example a neural network) is important. Then I will cover the different categories of state-of-the-art papers on this subject with a brief overview of the most well-known methods (such as LIME or SHAP) in each categories and a focus on the contributions made in my research team on the specific topic of time series classification. In particular, I will cover the following papers: Adversarial Regularization for Explainable-by-Design Time Series Classification - ICTAI 2020; A Performance-Explainability Framework to Benchmark Machine Learning Methods: Application to Multivariate Time Series Classifiers - IJCAI Workshop on Explainable Artificial Intelligence (XAI), 2020; XEM: An Explainable-by-Design Ensemble Method for Multivariate Time Series Classification, DAMI journal 2022; XCM: An Explainable Convolutional Neural Network for Multivariate Time Series Classification, Mathematics, 9(23), 3137).

Background Knowledge:

Advanced Knowledge about Machine Learning and Deep Learning

Bio: 

Elisa Fromont is a full professor at Université de Rennes France, since 2017 and a Junior member of the Institut Universitaire de France (IUF). She works at IRISA research institute in the INRIA LACODAM ("Large Scale Collaborative Data Mining") team. From 2008 until 2017, she was associate professor at Université Jean Monnet in Saint-Etienne, France. She worked at the Hubert Curien research institute in the Data Intelligence team. Elisa received her Research Habilitation (HDR) in December 2015 from the University of Saint-Etienne. Her research interests lie in (explainable) machine learning, data mining and, in particular, time series analysis.