ABSTRACT: Data scientists often train multiple models over and over to find the best hyperparameter combination for the best possible performance. However, once this initial exploration phase is complete, the continuous application of the trained model across different users and/or devices often requires much more energy than its initial training. In this talk, we will discuss how to make already ML models more resource-efficient once they have been trained.

BIOGRAPHY: Sebastian Buschjäger is a postdoctoral researcher and the Coordinator of Resource-Aware Machine Learning at the Lamarr Institute for Machine Learning and Artificial Intelligence in Dortmund since 2023. Before that, he was a PhD student at the Artificial Intelligence Unit of the TU Dortmund University in Germany where he finished his dissertation (Dr. rer. nat.) in 2022 with distinction. His research focuses on machine learning methods combined with small devices and specialized hardware.

WEBSITE: www.buschjaeger.it

ABSTRACT: Earth observation (EO) data archives are explosively growing as a result of advances in satellite systems. As an example, remote sensing (RS) images acquired by ESA’s Sentinel satellites (which are a part of EU’s Copernicus program) reach the scale of more than 10 TB per day. The “big EO data” is a great source for information discovery and extraction for monitoring Earth from space. Thus, accurate and scalable techniques for RS image understanding, search and retrieval have recently emerged.
This lecture will start by introducing the characteristics of the images acquired by sensors mounted on satellite platforms. Then, a general overview on scientific and practical problems related to RS image characterization, indexing and search from massive archives will be initially discussed, and the state-of-the-art methods and approaches that can overcome these problems will be presented. Particular attention will be given to deep hashing networks that learn a semantic-based metric space, while simultaneously producing binary hash codes for scalable and accurate content-based indexing and retrieval of satellite images. Practical applications will be provided throughout the lecture.

BIOGRAPHY: Begüm Demir is currently a Full Professor and the founder head of the Remote Sensing Image Analysis (RSiM) group at the Faculty of Electrical Engineering and Computer Science, TU Berlin and the head of the Big Data Analytics for Earth Observation research group at the Berlin Institute for the Foundations of Learning and Data (BIFOLD). Her research activities lie at the intersection of machine learning, remote sensing and signal processing. Specifically, she performs research in the field of processing and analysis of large-scale Earth observation data acquired by airborne and satellite-borne systems. She was awarded by the prestigious '2018 Early Career Award' by the IEEE Geoscience and Remote Sensing Society for her research contributions in machine learning for information retrieval in remote sensing. In 2018, she received a Starting Grant from the European Research Council (ERC) for her project "BigEarth: Accurate and Scalable Processing of Big Data in Earth Observation". She is an IEEE Senior Member and Fellow of European Lab for Learning and Intelligent Systems (ELLIS).

WEBSITE: www.bifold.berlin/people/prof-dr-beguem-demir.html

ABSTRACT: A lot of hype has evolved around generative machine learning models lately. Examples such as Stable Diffusion, Midjourney, DALL-E, and ChatGPT have gone viral and are used by many people in their everyday life. Less prominently known but similarly fascinating is the application of these powerful kind of models to the natural sciences. In this talk, we discuss how they can aid in the exploration of novel molecules and materials. To this end, we present the technical and conceptual challenges of molecule generation as well as the promises of advancements in this field, e.g. for renewable energies.

BIOGRAPHY: Niklas Gebauer is a doctoral researcher at the Machine Learning group of Prof. Dr. Klaus-Robert Müller at Technische Universität Berlin. He is a member of the BIFOLD graduate school and part of BASLEARN, the Berlin based joint lab of BASF and TU Berlin for machine learning. His work focuses on generative deep neural networks for molecular structures. He recieved his Master's and Bachelor's degrees in Computer Science from TU Berlin.

WEBSITE: www.bifold.berlin/people/niklas-gebauer.html

ABSTRACT: Modern machine learning models have revolutionized various fields such as image recognition, natural language processing, and speech recognition. However, training these models requires significant amounts of computational resources and energy, with the latter often coming from fossil fuel sources. The resulting carbon footprint raises growing concerns about the environmental impact of these models. As demand for machine learning models continues to grow, carbon-aware ML training will be critical to the sustainability of this field. In our talk, we will present recent research on carbon-aware and efficient machine learning that aims to mitigate these issues.

BIOGRAPHY: Dennis Grinwald is a PhD student at the Machine Learning Group at TU Berlin and member of the Graduate School at the Berlin Institute of the Foundations of Learning and Data (BIFOLD). His research focuses on efficient and robust ensemble learning on centralised and distributed data.

WEBSITE: www.bifold.berlin/people/dennis-grinwald.html

ABSTRACT: Decarbonizing global energy systems represents one of the most challenging task of our generation, playing a crucial role in mitigating the exhaustion of greenhouse gases on a global scale. The complexity of the task is amplified by the utility-scale implementation of sustainable generation units like wind turbines or photovoltaic power plants, which exhibit intermittent generation patterns. Additionally, the electrification of entire sectors such as heat supply and transportation further adds to the complexity of the problem. While machine learning (ML) methods can contribute to addressing these challenges, they often suffer from the perception of being opaque "black-boxes," leading to distrust and hindering their practical application. In this session, we will provide a concise overview of the domain of explainable AI (XAI) and explore how it can foster the development of more sustainable energy systems. Specifically, we will present application cases from the energy sector and highlight the advantages of employing XAI methods in this context.

BIOGRAPHY: Simon Letzgus is a doctoral researcher at the Machine Learning group of Prof. Dr. Klaus-Robert Müller at Technische Universität Berlin. After completing his master's degree in power engineering at Stuttgart University in 2015, he joined Siemens Energy to develop technical solutions related to the challenges of renewable energy systems. Since his return to academia, his research has focused on ML solutions for optimizing, monitoring and understanding complex technical systems with a specific emphasis on XAI for regression models.

WEBSITE: www.linkedin.com/in/simon-letzgus/

BIOGRAPHY: Volker Markl, Director of the BIFOLD Berlin Insitute, is a German computer scientist, database systems researcher, and a full professor. He leads the Chair of Database Systems and Information Management and is Director of the Berlin Institute for the Foundations of Learning and Data at the Technische Universität Berlin. He is also Chief Scientist and Head of the Intelligent Analytics for Massive Data Research Group at the German Research Center for Artificial Intelligence in Berlin. He has served as the President of the VLDB Endowment and has held an adjunct professorship at the University of Toronto. He currenty serves on the academic advisory council to the Alexander von Humboldt Institute for Internet and Society as well as the scientific advisory board to Software AG. Additionally, he co-chairs the Technological Enablers and Data Science Interdisciplinary Working Group of the ‘Plattform Lernende Systeme,‘ a platform of leading experts who are developing a roadmap for the responsible use of self-learning systems and AI, sponsored by BMBF, the German Federal Ministry of Education and Research. He is a strong proponent of data literacy, systems-oriented research, and computer science education in general.

MORE: www.bifold.berlin/people/prof-dr-volker-markl.html

BIOGRAPHY: Rainer Mühlhoff is leading the Ethics and critical theories of artificial intelligence research group at the Institute of Cognitive Science at the University of Osnabrück. He works on contemporary critical philosophy of the digital world; projects cover the ethics of AI, data protection and Big Data, digital technology and power, digital enlightenment, empowerment and participation.

WEBSITE: rainermuehlhoff.de/en/

BIOGRAPHY: Klaus-Robert Müller, Director of the BIFOLD Berlin Institute, received the Diploma degree in mathematical physics in 1989 and the Ph.D. in theoretical computer science in 1992, both from University of Karlsruhe, Germany. From 1992 to 1994 he worked as a Postdoctoral fellow at GMD FIRST, in Berlin where he started to built up the intelligent data analysis (IDA) group. From 1994 to 1995 he was a European Community STP Research Fellow at University of Tokyo in Prof. Amari’s Lab. From 1995 until 2008 he was head of department of the IDA group at GMD FIRST (since 2001 Fraunhofer FIRST) in Berlin and since 1999 he holds a joint associate Professor position of GMD and University of Potsdam. In 2003 he became a full professor at University of Potsdam, in 2006 he became chair of the machine learning department at TU Berlin. He has been lecturing at Humboldt University, Technical University Berlin and University of Potsdam. In 1999 he received the annual national prize for pattern recognition (Olympus Prize) awarded by the German pattern recognition society DAGM, in 2006 the SEL Alcatel communication award and in 2014 he was granted the Science Prize of Berlin awarded by the Governing Mayor of Berlin and in 2017 he received the Vodafone Innovations Award. Since 2012 he is Member of the German National Academy of Sciences Leopoldina and he holds a distinguished professorship at Korea University in Seoul. In 2017 he was elected member of the Berlin Brandenburg Academy of Sciences and also external scientific member of the Max Planck Society. For 5 years he was director of the Bernstein Center for Neurotechnology, from 2014 he became co-director of the Berlin Big Data Center (BBDC) and from 2018 simultaneously director of the Berlin Center for Machine Learning (BZML).

MORE: www.bifold.berlin/people/prof-dr-klaus-robert-mueller.html

ABSTRACT: The emerging field of Explainable AI (XAI) aims to bring transparency to today's powerful but opaque deep learning models. This talk will present Concept Relevance Propagation (CRP), a next-generation XAI technique which explains individual predictions in terms of localized and human-understandable concepts. Other than the related state-of-the-art, CRP not only identifies the relevant input dimensions (e.g., pixels in an image) but also provides deep insights into the model’s representation and the reasoning process. This makes CRP a perfect tool for AI-supported knowledge discovery in the sciences. In the talk we will demonstrate on multiple datasets, model architectures and application domains, that CRP-based analyses allow one to (1) gain insights into the representation and composition of concepts in the model as well as quantitatively investigate their role in prediction, (2) identify and counteract Clever Hans filters focusing on spurious correlations in the data, and (3) analyze whole concept subspaces and their contributions to fine-grained decision making. By lifting XAI to the concept level, CRP opens up a new way to analyze, debug and interact with ML models, which is of particular interest in safety-critical applications and the sciences.

BIOGRAPHY: Wojciech Samek is a professor in the Department of Electrical Engineering and Computer Science at the Technical University of Berlin and is jointly heading the Department of Artificial Intelligence at Fraunhofer Heinrich Hertz Institute (HHI), Berlin, Germany. He studied computer science at Humboldt University of Berlin, Heriot-Watt University and University of Edinburgh and received the Dr. rer. nat. degree with distinction (summa cum laude) from the Technical University of Berlin in 2014. During his studies he was awarded scholarships from the German Academic Scholarship Foundation and the DFG Research Training Group GRK 1589/1, and was a visiting researcher at NASA Ames Research Center, Mountain View, USA. Dr. Samek is associated faculty at the BIFOLD - Berlin Institute for the Foundation of Learning and Data, the ELLIS Unit Berlin, the DFG Research Unit DeSBi, and the DFG Graduate School BIOQIC, and member of the scientific advisory board of IDEAS NCBR - Polish Centre of Innovation in the Field of Artificial Intelligence. Furthermore, he is a senior editor of IEEE TNNLS, an editorial board member of Pattern Recognition, and an elected member of the IEEE MLSP Technical Committee and the Germany's Platform for Artificial Intelligence. He is recipient of multiple best paper awards, including the 2020 Pattern Recognition Best Paper Award and the 2022 Digital Signal Processing Best Paper Prize, and part of the expert group developing the ISO/IEC MPEG-17 NNC standard. He is the leading editor of the Springer book "Explainable AI: Interpreting, Explaining and Visualizing Deep Learning" (2019), co-editor of the open access Springer book “xxAI – Beyond explainable AI” (2022), and organizer of various special sessions, workshops and tutorials on topics such as explainable AI, neural network compression, and federated learning. Dr. Samek has co-authored more than 150 peer-reviewed journal and conference papers; some of them listed as ESI Hot (top 0.1%) or Highly Cited Papers (top 1%).

WEBSITE: https://iphome.hhi.de/samek

TBA

ABSTRACT: Neural networks have become a powerful tool to model potential energy surfaces and predict chemical properties. Starting from descriptor-based neural network approaches, this talk will focus on end-to-end learning of atomic representations and the incorporation of physical constraints. Finally, I will highlight some advanced applications of these techniques such as field-dependent potentials and generative neural networks.

BIOGRAPHY: Kristof T. Schütt received the PhD degree in machine learning at the Machine Learning Group of Technische Universität Berlin in 2018 for his research on deep learning for quantum chemistry. After developing neural networks for real-time speech enhancement in hearing aids at Audatic in 2020, he returned to TU Berlin as a group leader at the Berlin Institute for the Foundations of Learning and Data (BIFOLD) until 2022. Currently, he develops ML techniques for drug design at the Pfizer Machine Learning Research hub in Berlin. His research interests include interpretable neural networks, representation learning, generative models, and machine learning applications in chemistry and biology. 

WEBSITE: https://scholar.google.com/citations?user=0e49RfgAAAAJ&hl=en&oi=ao

ABSTRACT: Today, machine learning (ML) runs at various scales of hardware resources from the cloud and high-performance computing (HPC) centers to edge and Internet-of-Things (IoT) devices. To achieve resource-aware machine learning, we must understand the needs and challenges of ML applications at these different scales. In this talk, we will first investigate ways of improving hardware utilization on modern and powerful CPU-GPU co-processors, which serve as the commodity hardware for ML in the cloud and HPC, using workload collocation. Then, we will investigate performance and power trade-offs for ML-based image analysis in space using resource-constrained edge/IoT devices.

BIOGRAPHY: Pınar Tözün is an Associate Professor at IT University of Copenhagen. Before ITU, she was a research staff member at IBM Almaden Research Center. Prior to joining IBM, she received her PhD from EPFL. Her thesis received ACM SIGMOD Jim Gray Doctoral Dissertation Award Honorable Mention in 2016. Her research focuses on hardware-conscious machine learning, performance characterization of data-intensive systems, and scalability and efficiency of data-intensive systems on modern hardware.

WEBSITE: www.pinartozun.com 

ABSTRACT: Critical thinking has become more important than ever in the age of large ML models and Big Data. Its socio-technical context must be understood in order to develop common good oriented IT systems. In the workshop, we want to make use of collaborative public reasoning on this topic.

BIOGRAPHY: Stefan Ullrich is an Informatician with a minor degree in Philosophy who is researching on the impact of ubiquitous information technology systems on society. He is co-Chair of the Commission for Ethics in Research at the Technical University Berlin.

Website: www.weizenbaum-institut.de/en/spezialseiten/persons-details/p/stefan-ullrich/

ABSTRACT: Artificial intelligence and machine learning provide powerful tools for effective climate action and improving environmental sustainability. In various applications, such as energy, buildings, or transportation, they can help mitigate climate change by reducing greenhouse gas emissions. They can help adapt to a changing environment, monitor and slow the loss of biodiversity, and support the efficient use of resources. However, AI and ML are not a silver bullet and can only be part of the solution. This talk will provide an overview of the strengths and weaknesses of ML, some examples of applications and recurring themes, as well as trends in research and application in industry. Examples will be given from the Quantified Trees project on the use of AI for street tree irrigation in Berlin and from the BiGEye project where AI is used as part of a digital vegetation management to support the phase-out of glyphosate. 

BIOGRAPHY: Artificial intelligence and machine learning provide powerful tools for effective climate action and improving environmental sustainability. In various applications, such as energy, buildings, or transportation, they can help mitigate climate change by reducing greenhouse gas emissions. They can help adapt to a changing environment, monitor and slow the loss of biodiversity, and support the efficient use of resources. However, AI and ML are not a silver bullet and can only be part of the solution. This talk will provide an overview of the strengths and weaknesses of ML, some examples of applications and recurring themes, as well as trends in research and application in industry. Examples will be given from the Quantified Trees project on the use of AI for street tree irrigation in Berlin and from the BiGEye project where AI is used as part of a digital vegetation management to support the phase-out of glyphosate. 

WEBSITE: www.marcusvoss.com

ABSTRACT: Modern machine learning models have revolutionized various fields such as image recognition, natural language processing, and speech recognition. However, training these models requires significant amounts of computational resources and energy, with the latter often coming from fossil fuel sources. The resulting carbon footprint raises growing concerns about the environmental impact of these models. As demand for machine learning models continues to grow, carbon-aware ML training will be critical to the sustainability of this field. In our talk, we will present recent research on carbon-aware and efficient machine learning that aims to mitigate these issues.

BIOGRAPHY: Philipp Wiesner is a research associate and third year PhD student at TU Berlin, working in the research group on Distributed and Operating Systems (DOS) by Prof. Odej Kao. Before that, I’ve worked as a Software and Data Engineer at different companies in Berlin. My research focus is on carbon-aware computing: Aligning the power consumption of computing systems with the availability of renewable energy. Besides that, I’m interested in research on systems for machine learning as well as the co-simulation and testing of computing and energy systems.

WEBSITE: https://philippwiesner.org/

BIOGRAPHY: Herbert Zech, Director of the Weizenbaum Institute, habilitated in Bayreuth. From 2012, he worked at the University of Basel, initially as Extraordinarius for Private Law with a focus on Life Sciences Law, and since 2015 as Full Professor for Life Sciences Law and Intellectual Property Law. In April 2019, he moved to the Humboldt University of Berlin to the Chair of Civil Law, Technology and IT Law. The professorship is linked to the position of director at the Weizenbaum Institute for the Networked Society. Herbert Zech will be affiliated with the Institute for Computer Science of the Faculty of Mathematics and Natural Sciences in the form of a secondary membership. Herbert Zech also studied biology after his bar exam. Following on from this, he is particularly interested in the connection between the natural sciences and law. His main areas of expertise are technology law and intellectual property law. He is currently working on legal problems of Big Data and Artificial Intelligence, among others.

MORE: www.weizenbaum-institut.de/en/spezialseiten/persons-details/p/herbert-zech/