Seismologist and professor at the University Grenoble alpes

Short Bio

Michel Campillo, professor at the UGA, is a seismologist. He is therefore interested both in the processes responsible for earthquakes and in the seismic waves that reveal the structure of the Earth. He has studied large earthquakes but also the phenomena of transient deformation highlighted in recent years. He holds with Olivier Michel (Professor at Grenoble INP-UGA) the chair "Artificial Intelligence for natural hazard and geo-resources" at the Interdisciplinary Institute of Artificial Intelligence in Grenoble.

Presentation title

An example of the AI challenges for geophysics: unsupervised classification of large seismic and geodetic data sets.

Abstract

Recent discoveries in Earth sciences have been made possible with the analysis of large data sets, leading to an increase in observations. In a time when first order analyses reach their limits, huge data sets require new tools and new perspectives. We present approaches based on the scattering transform representation of the continuous geophysical time series from which several strategies of clustering can be tested, in a context of strongly unbalanced populations and various signal properties. We also investigate normality tests over multidimensional data without independence hypothesis to extract rare specific signals from long time series. We illustrate our approaches with the identification of population of signals and background noises in different geological settings.