Data Science and Statistics session speakers

We are pleased to confirm the following speakers in the Data Science and Statistics session, on June 1 from 16:15-17:45 CET:

Mia Hubert

Bio: Professor in Statistics and Data Science at the Department of Mathematics, KU Leuven

Title: Online teaching statistics for future data scientists

Abstract: For many years I have been teaching a master course on multivariate statistics and regression to students in Mathematics and Engineering. This year I developed a substantial make-over of the course. One reason was to make it better fit within the current online teaching world. But the main motivation was to direct it more towards data science. I will explain and show how I tried to combine theoretical concepts with applications and data visualisation.

John Chandler Johnson

Bio: Associate Professor at BI Norwegian Business School and Chief Scientific Officer at GAINSystems, Inc.

Title: Learning from Students: Executive Education Reveals Data Science Implementation Issues and Solutions

Abstract: Five years ago, we launched an executive education program with the objective to teach mid-career professionals enough data science to successfully collaborate with data science partners. We have trained more than 250 people in over 200 organizations throughout Europe, Asia, and North America. Engaging with experienced individuals in diverse organizations and industries reveals ubiquitous, non-modeling issues that consistently handicap data-driven projects. If domain experts understood these common issues and their implications, we anticipate that doomed projects would be abandoned earlier and ultimately successful projects would be completed more quickly. This paper discusses what we have learned from our students, and offers insights to help domain experts become more productive data science partners.

Ketil Malde

Bio: Researcher at the Institute for Marine Research and Associate Professor in Machine Learning at the Department of Informatics, University of Bergen

Title: The arduous voyage from accurate species recognition to sustainable marine resource management

Abstract: To address large-scale data challenges in marine science, we need to
automate data analysis and processing. There is now a large selection
of machine learning tools and methods that are readily available for
use. However, social and organizational obstacles hinder
effective deployment and thus progress towards the overall goal of
better marine resource management.

Geir Nilsen

Bio: PhD candidate at the Department of Mathematics, University of Bergen

Title: Epistemic Uncertainty Quantification in Deep Learning Classification by the Delta Method

Abstract: We propose a low cost variant of the Delta method applicable to L2-regularized deep neural networks based on the top K eigenpairs of the Fisher information matrix. A demonstration of the methodology is presented using a TensorFlow implementation, and we show that meaningful rankings of images based on predictive uncertainty can be obtained for two LeNet-based neural networks using the MNIST and CIFAR-10 datasets. Further, we observe that false positives have on average a higher predictive epistemic uncertainty than true positives.