Helmholtz Hosts D-G

Ansprechpartner

Jülich Research Centre (FZJ) - Institute of Neuroscience and Medicine Structural and functional organisation of the brain (INM-1) - Big Data Analytics Group (BDA)

Short summary of your group's research:  The Big Data Analytics group in Jülich develops software and methodology to build a microscopic resolution 3D model of the human brain, which includes detailed information about distributions of neuronal cells and nerve fibers, and microstructurally defined 3D maps of brain areas. This requires microscopic imaging and analysis of large quantities of histological brain sections at high throughput, leading to image datasets at the Petabyte scale.

At the intersection of neuroinformatics, computer vision and artificial intelligence (AI), our research addresses

• Data and workflow management for high throughput microscopic imaging,
• Machine learning and computer vision algorithms for biomedical image analysis on high performance computers,
• Software development for structured access to large image datasets and interactive 3D exploration of high-resolution brain atlases over the web.
•  

What infrastructure, programs and tools are used in your group?  The participant will be part of the local Helmholtz AI unit in Jülich. Our data management and image analysis workflows run on a distributed environment which includes Jülich’s modular high-performance computing systems. Image acquisition of whole-brain sections at micrometer resolution is performed on a high throughput microscopy facility. Outputs of our research are typically integrated with the human brain atlas hosted in the European research infrastructure EBRAINS, to which we contribute key developments. Most of our software is implemented in Python. Image analysis and machine learning approaches use common Deep Learning frameworks, typically PyTorch or Tensorflow.

What could a participant of the HIDA Mobility Program learn in your group? How could he or she support you in your group?

A participant in our group can learn about

• biomedical computer vision, especially processing and analysis of high resolution image data using state of the art AI methods
• distributed workflow management and parallel processing of big image data on supercomputers
• neuroinformatics.

They can support our group with practical experience in machine learning approaches for image analysis from the point of view of another application domain. They could also contribute with experience in inverse modelling, simulation, or computer graphics.

Ansprechpartner

Jülich Research Centre (FZJ) -  Simulation and Data Lab Quantum Materials, Jülich Supercomputing Centre, Institute for advanced Simulations

Short summary of your group's research

The Simulation and Data Laboratory Quantum Materials (SDLQM) serves as a dynamic hub for pioneering research at the intersection of simulation science and quantum materials. With a unique emphasis on leveraging High-Performance Computing and cutting-edge scientific Machine Learning, SDLQM offers a rich repository of expertise in quantum-based simulations within the realm of Materials Science. Our lab operates as a catalyst for innovation, actively engaging in dedicated projects and hosting research initiatives that delve into the essential domains of method refinement, algorithmic enhancement, and performance optimization. At the forefront of simulation science, SDLQM thrives on fostering cross-disciplinary collaborations spanning the global landscape. We take pride in our robust engagement within both the European and Japanese research arenas. Within the EU, our lab spearheads pioneering research and development initiatives, propelling the domain of Materials for Energy to new horizons. We are resolutely focused on facilitating exascale simulations of quantum transport, unraveling complexities that hold the key to transformative advancements. Our endeavors transcend borders, as we cultivate robust ties with esteemed partners in Japan, notably nurturing a dynamic synergy with the illustrious RIKEN supercomputing center in Kobe.

What infrastructure, programs and tools are used in your group?

Guided by an unwavering commitment to excellence, SDLQM not only facilitates extensive-scale simulations on state-of-the-art HPC platforms but also pioneers the training of intricate learning models driven by insights garnered from in-silico simulations. SDLQM operates at the heart of the Juelich supercomputing center, a pioneering institute that presently houses one of the largest supercomputers across the European Union. This convergence of minds and machines is set to elevate to unprecedented heights, with Europe's inaugural exascale supercomputer slated for 2024. Immerse yourself in a realm of unbounded possibilities, as the lab cultivates expertise encompassing high-performance computing, parallel programming, numerical linear algebra, quantum transport simulation, functional Renormalization Group methods and Tensor Networks. As a nexus at the frontier of knowledge, we seamlessly unite the realm of materials science simulation with the realm of massively parallel machines, thereby nurturing a community driven by cutting-edge innovation and collaborative excellence.

What could a guest researcher learn in your group? How could he or she support you in your group?

The doctoral researcher will be granted privileged access to JSC's cutting-edge supercomputing hardware, ushering them into the realm of computational supremacy. Working in close tandem with the group's accomplished experts, they will embark on a transformative journey of code development, adaptation, and optimization. Moreover, the doctoral researcher will gain firsthand exposure to the dynamic intersection of numerical linear algebra and materials science simulation. This immersive journey will encompass code integration, delving into the intricacies of parallel programming best practices, and honing indispensable software management skills. Join us in shaping the future of quantum-centric materials research, where innovative simulations and data-driven discoveries converge.

Ansprechpartner

CISPA - Helmholtz Center for Information Security - SprintML

Three-sentence summary of your group's research: Our research is focused on secure and trustworthy Machine Learning as a Service (MLaaS). We design robust and reliable machine learning methods for training and inference of ML models while preserving data privacy and model confidentiality. We are focused on the state-of-the-art ML methods and models, such as, self-supervised learning, large language and vision models (LLMs), foundational models.

What infrastructure, programs and tools are used in your group? We use large scale GPU clusters and program in the latest ML frameworks.

What could a guest researcher learn in your group? How could he or she support you in your group? As a guest researcher you can learn how to use tools from data privacy and model confidentiality to enable trustworthy machine learning. We also support our guests with theoretical analysis as well as running large scale experiments to publish at the top ML conferences.

Ansprechpartner

Karlsruhe Institute of Technology (KIT) -  Institute for Water and Environment - Hydrology

Short summary of your group's research: Our group has a wide range of hydrology-related interests comprising conceptual and physically based hydrological modeling, ecohydrology, application of thermodynamic principles to hydrology, data-infrastructures and data-based learning. In the context of HIDA, especially the latter topic is of interest: My colleague Ralf Loritz and I focus on applying machine-learning approaches to hydrological modeling,  and applying concepts from information theory for both hydrological system analysis and as a basis for building hybrid hydrological models merging information from data and knowledge.

What infrastructure, programs and tools are used in your group? We have developed and use a range of open-source Matlab- and Python-based tools for analysis and modeling in hydrological and geostatistical contexts. See more on Link.

What could a participant of the HIDA Mobility Program learn in your group? How could he or she support you in your group? 

What you can learn from us:

• Applying state-of-the-art tools for supervised and unsupervised learing in hydrological settings
• Applying state-of-the-art machine-learning tools for hydrological forecasting and prediction
• The beauty and generality of concepts from information theory, and how they can be used for hydrological analysis and model building

What we can learn from you:

• A fresh look and feedback on what we do
• New insights into our methods by applying them to your data and problems
• Learn about alternative methods

Ansprechpartner

Deutsches Elektronen-Synchrotron DESY - Machine beam control (MSK)

Short summary of your group's research: 

Intelligent Process Controls (IPC) is a subgroup of the Machine Beam Controls (MSK) group at DESY, pushing forward innovative research for the autonomous operation of particle accelerators at the interface of machine learning, control. For this using reinforcement learning and other cutting-edge optimization techniques. IPC is also engaged in developing advanced feedbacks and enhancing fault diagnosis and anomaly detection through machine learning.

What infrastructure, programs and tools are used in your group?

The participant will have access to high-performance computing infrastructure at DESY. As programming languages, we mainly use Python.

What could a guest researcher learn in your group? How could he or she support you in your group?

As an interdisciplinary research team of experts from control theory, computer science and physics, a participant of the HIDA Mobility Program can gain experience in different directions and support many different applied projects. Here data mining projects as for anomaly detection are possible but also control and optimization problems. For the latter, we strongly encourage the guest researcher to participate in shifts and applying the developed methods to the accelerator. 

Ansprechpartner

Jülich Research Centre (FZJ) - Brain and Behaviour (INM-7)

Short summary of your group's research: We develop and apply novel methods to better understand  the organization of the human brain and to relate it to behavioral phenotypes in healthy and as biomarkers in clinical populations. To do so, we capitalize on machine learning approaches. In addition to these scientific objectives, we also develop open access methods and tools for data management and reproducible research.         

What infrastructure, programs and tools are used in your group? Python, R, MATLAB, HPC, machine learning, neuroimaging

What could a participant of the HIDA Mobility Program learn in your group? How could he or she support you in your group?  A participant can choose from diverse topics from machine learning, data management to biomarker discovery. For technically oriented participants we offer opportunities to contribute to our growing set of tools. For application-oriented guests we offer participation in ongoing projects to uncover brain-behavior relationships.

Ansprechpartner

Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research -  Ecosystem Data Science Group

Three-sentence summary of your group's research: Our research program focuses on understanding the ecology and evolution of naturally occurring microbial life using integrated ‘omics strategies and laboratory experiments. Our group aims to develop new computational approaches and create advanced software platforms that intend to generate hypotheses from complex environmental data to bring us closer to explain mechanisms by which microbes interact with their surroundings, evolve, disperse, and initiate or adapt to environmental change. We are environment agnostic as microbial ecology and evolution happens everywhere, thus, our recent work ranges from the human gastrointestinal tract and oral cavity to sewage systems, and insect ovaries. That said, we find marine systems much more intriguing than most other habitats and desire to channel our future efforts to understanding microbial life in sunlit and dark oceans.

What infrastructure, programs and tools are used in your group? Having to work with extremely large and complex 'omics datasets, the initial stages of our data analyses typically rely on high performance computing infrastructures. Although, as we are often much more interested in targeted analyses of subtle phenomena that reveal specific insights for specific questions rather than overall analyses of our data, later stages of our data analyses often rely on R and/or Python programs and interactive visualization solutions.

Our group includes many of the key developers of anvi'o (https://anvio.org/), a comprehensive open-source software platform that brings together many aspects of today’s cutting-edge computational strategies of data-enabled microbiology, including genomics, metagenomics, metatranscriptomics, pangenomics, metapangenomics, phylogenomics, and microbial population genetics in an integrated and easy-to-use fashion through extensive interactive visualization capabilities.

Anvi'o stands on more than 200,000 lines of Python and JavaScript code that follow modern software development paradigms thanks to the voluntary contributions of more than 30 developers from the academia and industry to empower microbiologists for their complex needs through an architecture that enables reproducible science and simple interfaces for data access.

What could a guest researcher learn in your group? How could he or she support you in your group? Our group can help addressing novel questions through complex 'omics datasets and ad hoc software development, support others with insights from integrated 'omics strategies, and collaborate on integrating software solutions into the anvi'o software ecosystem to make them available to the broader community of life scientists.

Ansprechpartner

Helmholtz-Zentrum Hereon - Institute of Surface Science

Short summary of your group's research: The research activities at the Institute of Surface Science revolve around the development of ultra‐light‐weight materials and surface protection technologies for the application areas air and ground transportation, infrastructures, regenerative medicine, as well as energy storage and functional materials. Gaining control over surface reactivity characteristics and a reliable prediction of service-life aspects of the employed materials are prerequisites to unlock their full potential in the respective application areas. In this context, we utilise information derived from multiscale simulations, experimental investigations in labs and field data from industrial end-users to gain a comprehensive insight into the material behaviour and employ a variety of supervised and unsupervised machine learning techniques for the discovery of effective additives prior to experimental testing as well as to predict how materials and structures behave during their service-life after exposure to the environment with focus on a sustainable circular economy.

What infrastructure, programs and tools are used in your group? Access to Hereon’s HPC Cluster, HPC workstations in the group for less demanding jobs, Several simulation software packages (DFT, MD, FEM) and tools to generate molecular descriptors (alvaDesc, rdkit), Python-based machine learning workflows.

What could a participant of the HIDA Mobility Program learn in your group? How could he or she support you in your group?  

Our offer:

• Insights into material science with focus on materials degradation and corresponding corrosion mitigation strategies.
• Techniques to generate molecular descriptors for quantitative structure-activity/property relationship models and approaches to identify features that are relevant to the chosen target property as well as subsequent model development.
• Opportunity to validate the robustness of your machine learning model(s) directly with our experimentalists who provide the training data.  

Your support:

• You could either apply your machine learning knowledge to our existing training data or shape a new project together with our interdisciplinary team with focus on corrosion mitigation and/or predictive maintenance.
• You could employ your experience in materials modelling to generate additional input features for our predictive models.
• You could help us to develop tools that can be utilised by end-users without any domain expertise

Ansprechpartner

Jülich Research Centre (FZJ) - Juelich Plant Phenotyping Center/Institute of Bio- and Geosciences IBG2: Plant Sciences

Short summary of your group's research: We focus on the application and development of high-throughput, non-invasive phenotyping methods for screening of plant shoot and root productivity traits. In our group teams of plant biologists, engineers and image processing specialists cooperate to realize screening methodologies for shoot and root growth and architecture combined with physiological measurements of plant performance and plasticity to environmental challenges. Interpretation of results requires the evaluation of multi-dimensional data structures which include dynamic changes of plant phenotypes over time, environmental and imaging sensors data.

What infrastructure, programs and tools are used in your group? We have developed and operate a unique set of automated experimental platforms to analyze plant phenotypes in a minimally-invasive fashion, please click here to learn more.

What could a participant of the HIDA Mobility Program learn in your group? How could he or she support you in your group?  Participants of the HIDA Mobility Programwould benefit from learning: a) methodologies for high-throughput screening in plants; b) data analyses and visualization of complex data sets; c) experimental designs for plant phenotyping. Participants would support our continued efforts in improving and automating data analyses using existing data stored in several institutional databases. 

Ansprechpartner

Helmholtz Munich – ELLIS lab

Three-sentence summary of your group's research: Our group focuses on the interface between Bayesian inference and deep learning with the goals of improving robustness, data-efficiency, and uncertainty estimation in these modern machine learning approaches. While deep learning often leads to impressive performance in many applications, it can be over-confident in its predictions and require large datasets to train. Especially in scientific applications, where training data is scarce and detailed prior knowledge is available, insights from Bayesian statistics can be used to drastically improve these models. Important research questions include how to effectively specify priors in deep Bayesian models, how to harness unlabeled data to learn re-usable representations, how to transfer knowledge between tasks using meta-learning, and how to guarantee generalization performance using PAC-Bayesian bounds.

What infrastructure, programs and tools are used in your group? We make use of high-performance CPU and GPU compute clusters (via Slurm) and regularly use deep learning frameworks (mostly PyTorch and JAX) together with the standard Python data analysis and visualization stack (scikit-learn, pandas, numpy, matplotlib, etc). On the theory side, we are interested Bayesian model selection, approximate Bayesian inference, and PAC-Bayesian generalization bounds.

What could a guest researcher learn in your group? How could he or she support you in your group? You can learn about the latest and greatest in Bayesian machine learning as well as how to train both small and (very) large deep learning models on one of the largest GPU clusters in Europe. You can apply cutting-edge ML models to a range of scientific data to push the boundaries of ML-powered scientific discovery, drug development, and understanding of health and disease. You can experience working directly at the intersection of theory and application with lots of freedom to explore either direction to suit your interests.

Ansprechpartner

Karlsruhe Institute of Technology (KIT)- Computational Science & Mathematical Methods

Short summary of your group's research: The research group Computational Science and Mathematical Methods (CSMM) under the supervision of Prof. Dr. Martin Frank is an interdisciplinary research group working on different challenges in mathematical modeling.  We are unified by our interest in method-oriented mathematics, in mathematical modeling inspired by applications, and in the didactics of mathematical modeling. Our research focuses on bringing modern mathematical techniques such as modeling, simulation, optimization, inverse problems, uncertainty quantification, and machine learning/artificial intelligence into the real-world. Furthermore, we take an interest in kinetic theory, developing models, numerical methods and software implementations and combining them with the aforementioned techniques.

What infrastructure, programs and tools are used in your group? Among other things, we use the modern high performance computing infrastructure of KIT, namely bwUniCluster.  We develop software and tools in Julia, Python, C++ and Matlab, depending on the project. We employ several project specific software stacks for uncertainty quantification and medical image processing.

What could a participant of the HIDA Mobility Program learn in your group? How could he or she support you in your group? As an interdisciplinary research group, we offer a lot of possibilities for a participant of the HIDA Mobility Program and many different interesting projects to support. A mathematically oriented participant could learn interesting aspects about machine learning based numerical methods for kinetic equations. Furthermore, the group offers projects in mathematical methods for uncertainty quantification and dynamical low rank.

An application-oriented participant may find interest in the project about neural network based medical image processing or uncertainty quantification in radiation therapy.

A participant, who is interested in teaching, can learn about bringing modern and method oriented mathematics to the classroom and can work on applied mathematical projects with high school students.

We develop an open source simulation toolkit for radiation therapy planning, which also offers the possibility to engage and participate. Detailed descriptions of our projects can be found on our group homepage.

Ansprechpartner

Karlsruhe Institute of Technology (KIT) - Artificial Intelligence for materials science (AiMat), Department of Informatics

Short summary of your group's research: The AiMat group works on the development and application of AI and machine learning (ML) methods for materials science and chemistry. Focus topics are (graph) neural networks for the prediction of molecular and materials properties, the combination of interpretable AI methods, ML-enhanced simulations and automated experiments to autonomous materials acceleration platforms, and generative models for inverse materials design.

What infrastructure, programs and tools are used in your group? 

Infrastructure: HPC systems as BWUniCluster and HoreKa
Programs: Python and machine learning libraries, as well as quantum chemistry simulations

What could a participant of the HIDA Mobility Program learn in your group? How could he or she support you in your group?  We welcome participants with either a background in machine learning and methods development, who are interested in applying the models to interesting scientific questions, or participants with a background in materials science/chemistry/physics, who want to learn more about machine learning and teach us about their application areas!

Ansprechpartner

GFZ Helmholtz Centre for Geosciences - Hydrothermal Systems

Short summary of your group's research: Our working group explores the Earth’s thermal field and geothermal resources, studies involved processes, quantifies their relevance, and provides knowledge on its behavior over time and across scales. The group is aimed at identifying priority targets for different geoenergy utilizations. With our work, we contribute to the transformation of the conventional energy system and to the reduction of CO2 emission.

With an applied focus, we elaborate exploration strategies to advance the successful development of geoenergy utilization in urban areas, where conventional surface exploration methods are often not applicable. We develop advanced methodologies for cross-scale characterization and for a better scale-dependent parameterization of the subsurface for risk reduction. A key concept is to combine geological expertise with a multi-methodological approach to establish adequate and reliable conceptual subsurface models. Fields of application include the sustainable provision of geothermal energy or the successful application of (thermal) storage systems.

With a more fundamental perspective, we investigate thermal processes in the crust and analyze the thermal field (including heat flow), and provide boundary conditions for multi-process integrated geodynamic models. With the integration of multidisciplinary observation data, we improve the understanding of the present thermal state and involved processes in subsurface thermal geosystems: from local to global views, from rock to lithospheric scales and across time domains. 

What could a participant of the HIDA Mobility Program learn in your group? How could he or she support you in your group? We welcome all researchers with interest in geothermal field exploration, numerical methods and related data science. A guest researcher could learn about thermal field modeling approaches and their applications. Possible joint projects could involve processing and mapping of regional to global heat flow data, conducting analysis of observational data for mapping applications.

Prof. Dr. Julien Gagneur

Helmholtz Munich  – Computational Health Center

Three-sentence summary of your group's research: Our goal is an improved understanding of the genetic basis of gene regulation and its implication in diseases. To this end, we employ AI and statistical modeling of 'omic data and work in close collaboration with experimentalists. We are both based at the TUM and HMGU.

What infrastructure, programs and tools are used in your group? We work with python, ML libraries (pytorch, etc.) and R/Bioconductor.
We have developed multiple software for regulatory genomics and genetic diagnostics including :

• API and repos for trained ML models in genomics Kipoi
• sequence-based model of splicing MMSplice (GitHub)
• Base pair resolution modeling of genome-wide assays BPnet (Article regarding the topic)
• Omics based diagnosis of rare disorders: OUTRIDER, FRASER, DROP: (Protocol regarding the topic)  

Ansprechpartner

German Aerospace Center (DLR) - Visual Computing and Engineering Department

Short summary of your group's research: The mission of the DLR Institute for Software Technology is research and development in software engineering technologies, and the incorporation of these technologies into DLR software projects. The department of Software for Space Systems and Interactive Visualization investigates and develops methods for robust and reusable software solutions for the design and operation of space missions as well as for interactive visualization applications.

What infrastructure, programs and tools are used in your group? Besides modern computer systems, we are using well-equiped laboratories for flight software integration into onboard sytems, for the investigation of immersive software in virtual environments, and for evaluation of software solutions into mixed and extended reality.  This includes large immersive virtual reality powerwall displays, multi-pipe multi-touch display walls, and fully embedded systems development equipement. Additionally, we are operating a software laboratory for highly agile joint software development sessions. All systems have access to institute server for high-performance computing, cluster solution development, and high-end visualization approaches.

What could a participant of the HIDA Mobility Program learn in your group? How could he or she support you in your group? We are looking for excellent scientists for all activities in the research groups of the department: robust and resilient onboard software, formal verification of spacecraft design, digital transformation in the space domain, in-situ data processing and visualization on high-performance computers, topology-based data analysis for large-scale datasets, visual data analytics, virtual reality, mixed reality, extended reality, human factors and user studies, applications in DLR's domain like space, aeronautics, transportation, climate change, and more. 

Ansprechpartner

Helmholtz-Zentrum Dresden-Rossendorf (HZDR) - Exploration and characterization

Short summary of your group's research: We are developing imaging technologies for the characterization of complex material streams and the Earth's surface. Our approach is based on sensor integration and processing using machine learning. Our platforms include drones, robots and conveyor systems.

What infrastructure, programs and tools are used in your group? We have our own imaging sensors, a GPU HPC and develop our own python tools.

What could a participant of the HIDA Mobility Program learn in your group? How could he or she support you in your group? A guest researcher would learn to work with cutting edge technology and process data with the latest developments in computer vision and machine learning.