Solving large-scale inverse problems in Python with PyLops

Tutorial

Go to NumFOCUS academy page.

Inverse problems are at the core of many scientific disciplines. When working with large data and/or model vectors, an explicit matrix representation of the problem becomes unfeasible. In this tutorial we present PyLops, an open-source Python library that uses linear operators to solve matrix-free inverse problems. Examples of image deblurring and CT reconstruction will be presented as use cases.

Speakers

Matteo Ravasi

I am a geophysicist with a formal education in Telecommunication Engineering and strong background in Applied Mathematics, Physics, Statistics, and Computer Science. I enjoy tackling the entire life-cycle of complex scientific problems from developing new theoretical formulations and implementing numerical solutions, to creating user-friendly software libraries and front-end interfaces.

My expertise includes model- and data-driven inverse problems, statistical signal processing, wave propagation, uncertainty analysis, seismic imaging, quantitative interpretation, 4D seismic, multicomponent acquisition and processing, and numerical modelling. I am also a keen open-source contributor and I am currently involved in the development of several Python libraries such as PyLops and spgl1.

Ivan Vasconcelos

I am an Assistant Professor of Applied Geoscience at Utrecht University. My research interests broadly revolve around wave & diffusion phenomena, inverse problems and imaging science. Together with a talented team, we work on seismic and radar imaging for Earth and planetary studies, experimental rock & wave physics, material science in imaging problems, and connections between geophysical and medical imaging.

Our work focuses on theoretical and experimental fundamentals - but we have a strong belief in developing robust, reproducible, user-friendly and ready-to-use open-source software. We are users, developers and contributors of the Pylops package, and currently working toward a new package for Material Science applications. Our door is always open to collaborations on software for research, applications, teaching or just fun.

David Vargas

I am a PhD student at Utrecht University, The Netherlands. I am currently working for the “Utrecht Consortium for Subsurface Imaging”. My formal education is in Physics Engineering from the National University of Colombian and holds a master’s in Geophysics from LMU-Munich. My research interests focus on the development of novel data-driven seismic redatuming technics and aim at finding ways to apply Marchenko Imaging in real scenarios by bridging the gaps between the theoretical developments and the specific case of study in real scenarios. Some of the applications include time-lapse Marchenko imaging for reservoir monitoring and characterization as well as waveform inversion of redatumed fields for sub-salt studies. I use the Pylops package to perform the numerically intensive computations driven my research.