The Helmholtz AI Annual Retrospective and Award Ceremony showcased researcher’s brilliant work throughout the year as participants joined in to commemorate the outstanding contributions to the field. The Helmholtz AI Awards spotlighted and celebrated the achievements of AI researchers within the Helmholtz Association in three categories...
As yet another transformative year in the realm of artificial intelligence for science comes to an end, the Helmholtz AI community came together last week to mark the culmination of achievements, celebrate the collective strides made, and set the stage for an even more promising upcoming year.
The Helmholtz AI Annual Retrospective and Award Ceremony showcased researcher’s brilliant work throughout the year as participants joined in to commemorate the outstanding contributions to the field. The Helmholtz AI Awards spotlighted and celebrated the achievements of AI researchers within the Helmholtz Association in three categories:
- Best paper.
- Best digital resource.
- Best PhD dissertation.
Let us introduce the awardees – congratulations to them all!
CATEGORY BEST PAPER:
“Active learning-assisted neutron spectroscopy with log-Gaussian processes”
Mario Teixeira Parente
Interim Professor for Computational Statistics & Data Science at the Department of Statistics, Ludwig-Maximilians-Universität München
*Received for project done while with the Jülich Centre for Neutron Science (JCNS) at Heinz Maier-Leibnitz Zentrum (MLZ) at a Post-doc position.
Research interests: uncertainty quantification, Bayesian inverse problems, gaussian process regression.
“At large-scale neutron research facilities, the first hours of an experiment using a three-axes spectrometer (TAS) are crucial for its overall success. We developed an approach, ARIANE, that can assist experimenters by making autonomous decisions in this first part of the experiment and providing guidance for more detailed investigations of a material which require human expertise and experience. For this, ARIANE makes advanced use of Gaussian process regression, a well-known statistical technique, and log-normal distributions, as both can be deeply connected with scenarios at a TAS. Indeed, ARIANE has already demonstrated its value several times in real TAS experiments at renowned European neutron sources.
My team and I are very delighted to receive the "Best Paper 2023" award, as it is a great recognition of our dedicated work over the last three years.”
CATEGORY BEST DIGITAL RESOURCE:
Metrics Reloaded: A framework for problem-aware metric recommendations for image analysis
Annika Reinke (*on behalf of the metrics consortium)
Div. Intelligent Medical Systems
German Cancer Research Center (DKFZ); PostDoc, Group Lead and Deputy Head
Research interests: validation, challenges, good scientific practice, biomedical image analysis
“Validation metrics are key for the reliable tracking of scientific progress and for bridging the current chasm between artificial intelligence (AI) research and its translation into practice. However, increasing evidence shows that particularly in image analysis, metrics are often chosen inadequately in relation to the underlying research problem. With Metrics Reloaded, an international multi-centered initiative with more than 70 researchers involved, we developed the first comprehensive metric recommendation framework guiding researchers in the problem-aware selection of validation metrics for machine learning algorithm validation.”
CATEGORY BEST PHD DISSERTATION:
“Deep neural networks for large-scale cytoarchitectonic mapping of the human brain”
Institute of Neuroscience and Medicine (INM-1), Research Centre Jülich and Helmholtz AI
PostDoc, Helmholtz AI Young Investigator Group Leader "Large-scale AI for brain mapping"
Research interests: representation learning, geometric deep learning, microscopic image analysis, high-performance computing, visualization
“In my PhD thesis “Deep Neural Networks for Large-Scale Cytoarchitectonic Mapping of the Human Brain”, I developed methods for the automatic characterization and classification of cytoarchitectonic areas in the human brain based on high-resolution microscopic scans of histological brain sections. Analyzing cytoarchitecture (i.e., the spatial organization of neuronal cells) enables the segregation of the human brain into architecturally distinct areas. These cytoarchitectonic areas are crucial to identify functional, genetic, physiological, molecular, and connectivity data with microstructurally well-defined entities. As such, they are fundamental building blocks for the development of high-resolution multi-modal human brain atlases. In my thesis, I developed deep learning methods that for the first time facilitate cytoarchitectonic analysis in terabyte to petabyte-scale microscopic image datasets, including strategies for efficient learning from limited labelled data, representation learning approaches, and methods for joined learning from images and geometric data.”