Fermer

Przemyslaw Juda

PhD Student

 

Research:

My main focus are Monte Carlo samplers and Multiple-point Statistics for solving inverse problem in hydrogeology. I have been investigating different strategies to improve convergence of inverse methods. In the first part of my thesis, I worked on the importance of prior selection for probabilistic inversion. In the second part, I have been looking into techniques for speeding up the inverse algorithms, mainly with machine learning.

My cross-validation framework for choosing simulation parameters adressed the problem of tuning stochastic categorical geostatistical models to observed data (such as geological facies). We demonstated that it is especially applicable for multiple-point statistical simulations (including non-stationary cases).

Another study showed that machine learning has potential to speed-up inversion algorithms by applying classification before running the forward solver (often computationally expensive). We found that ensemble methods, such as random forests or AdaBoost, outperform convolutional neural networks if training data is relatively scarce (500 examples).

Currently, I am working on a synthetic pumping test set-up which aims to compare different inversion algorithms: posterior population expansion, iterative ensemble smoother and a Monte Carlo solver  coupled with generative adversarial network (GAN).

 

Main publications:

  • Juda, Przemysław, and Philippe Renard. “An Attempt to Boost Posterior Population Expansion Using Fast Machine Learning Algorithms.” Frontiers in Artificial Intelligence 4 (2021). https://doi.org/10.3389/frai.2021.624629.
  • Juda, Przemysław, Philippe Renard, and Julien Straubhaar. “A Framework for the Cross-Validation of Categorical Geostatistical Simulations.” Earth and Space Science 7, no. 8 (2020): e2020EA001152. https://doi.org/10.1029/2020EA001152.
  • Dagasan, Yasin, Przemysław Juda, and Philippe Renard. “Using Generative Adversarial Networks as a Fast Forward Operator for Hydrogeological Inverse Problems.” Groundwater 58, no. 6 (2020): 938–50. https://doi.org/10.1111/gwat.13005.

 

Teaching Responsabilities:

  • Spring 2019: Teaching assistant, TP Expériences et laboratoire en géologie appliquée
  • Fall 2018: Teaching assistant, Géostatistique et modélisation inverse
  • Spring 2018: Teaching assistant, TP Expériences et laboratoire en géologie appliquée

 

Education:

  • 2015 – 2017: MSc in Computational Science and Engineering, Ecole Polytechnique Fédérale de Lausanne (EPFL), Switzerland
  • 2010 – 2015: Engineer in Technical Physics, Wrocław University of Science and Technology, Poland

 

Positions held:

  • 02.2018 – Present: Doctoral Assistant, Université de Neuchâtel, Switzerland
  • 11.2017 – 01.2018: Scientific Trainee in Laboratory of Computational Science and Engineering COSMO, Ecole Polytechnique Fédérale de Lausanne (EPFL), Switzerland
  • 09.2013 – 09.2014: Technical Student (intern) in Technology Department at CERN (European Organization for Nuclear Research), Geneva, Switzerland

Contact

Office: E321

​Phone: +41 (0)32 718 26 86

 

https://github.com/pjuda

 

Centre d'hydrogéologie et géothermie (CHYN)

Emile-Argand 11

CH-2000 Neuchâtel