Jan Byška is an Assistant professor at Masaryk University, Czech Republic, and an Adjunct Associate professor at the University of Bergen, Norway. He received his Ph.D. degree in 2016 from Masaryk University. From 2016 to 2019, Dr. Byška worked as a postdoc at the University of Bergen.
Dr. Byška’s work centers around visual data science, focusing on various challenges in visualizing biomolecular and spatio-temporal data. He coauthored multiple software tools for visualization of molecular data, among the most notable CAVER Analyst. Additionally, his research interest recently extended towards virtual reality and its use for the exploration of complex molecular data. He is also actively participating in the organization of multiple scientific events in the area of biomolecular visualization (BioVis, MolVA).
Title: Visualizing Large Biomolecular Data
Abstract: With recent technological advancements, we can simulate or gather extensive datasets through experimentation. While this available data opens doors to unprecedented discoveries, it also presents notable hurdles in subsequent analysis. In this invited talk, I will discuss the difficulties of working with extensive biomolecular datasets, and I will provide instances illustrating how the field of visual data science can effectively address these issues. The presentation will primarily center around extensive spatio-temporal data such as molecular simulations consisting of millions of timesteps or larger drug design datasets for virtual screening.
Vaneeda Allken is a machine learning researcher at the Institute of Marine Research. Her focus is on developing practical machine-learning based solutions for problems in marine science. Her latest work aims at automating fish counting and the estimation of fish species distribution using images obtained from trawl surveys.
Title: Bridging the gap between theory and practice AI
Abstract: Although machine learning models are getting increasingly sophisticated and becoming more widely accessible to the general public, many enterprises still struggle to put these methods to practical use. One prominent cause of this is that the impressive performance obtained on benchmark data often fails to translate to real-world use cases. It is challenging to build sufficiently representative datasets from scratch. Developing an effective model also requires a clear understanding of the objective and the manner in which the model will be deployed. Proper communication with the domain experts and the users is important. In this talk, I will discuss some of the roadblocks we faced while developing and implementing machine learning models in a practical setting, possible solutions and suggestions for the way forward.
Mattias Villani is a Professor at the Department of Statistics at Stockholm University, Sweden. He has previously held positions at the Central bank of Sweden and Linköping University, where he built up and headed the Division of Statistics and Machine Learning. His research centers on computationally efficient methods for Bayesian inference, prediction and decision making, using flexible probabilistic models. The methods are applied to a wide range of problems in economics, neuroimaging, transportation, robotics and text analysis.
Title: An Introduction to Bayesian Learning for Uncertainty Quantification and Decision Making
Abstract: Data science and machine learning is inherently about making decisions under uncertainty. In this lecture I will explain how the Bayesian approach gives a natural quantification of uncertainty, in a form that is directly useful for decision making. I will also briefly discuss how computer simulation is used as an everyday tool for Bayesian computations, and introduce a couple of probabilistic programming languages that can be used to perform a Bayesian analysis with minimal effort.
Stein Andreas Bethuelsen is an Assoc. Prof. in the Department of Mathematics at UiB. He is a probabilist working on unraveling the theoretical underpinnings of stochastic processes with application areas e.g. in physics and biology. A current focus is the study of random motion in random media, interacting particle systems, chains of infinite order and Gibbs measures.
Title: Markov chains and applications to data science
Abstract: In this talk I will briefly survey the basic theory of Markov chains and illustrate some of its fascinating applications to data science (such as the Markov chain Monte Carlo method).
Cagatay Turkay is a Professor at the Centre for Interdisciplinary Methodologies at the University of Warwick, UK and a Turing Fellow at the Alan Turing Institute, London, UK. His research investigates the interactions between data, algorithms and people, and explores the role of interactive visualization and other interaction mediums such as natural language at this intersection. He designs techniques and algorithms that are sensitive to their users in various decision-making scenarios involving primarily high-dimensional and spatio-temporal phenomena, and develops methods to study how people and society work interactively with data and computed artefacts.
Title: Doing Visual Data Science — Foundations, Techniques and Practice
Abstract: In this part of the summer school, we will be opening up the concept of visual data science as an analytical methodology, as a way of thinking, and a way of doing data science. Data visualizations — including those that are interactive — offer effective, rich, robust and critical ways of making inferences from data, working with statistical and machine learning models, as well as providing a bridge from analysis to dissemination and reproducibility. This course will explore the theoretical foundations of visual data science and discuss visual thinking and interactive visualization methods and techniques that underpin the visual data science process. We will explore these topics through reflective activities and examples where visual data science approaches have been adopted. The applied part of the course will explore the practice of visual data science through coding a number of techniques using Python libraries, and a range of hands-on activities will develop critical thinking skills in designing effective visual data science processes. The students will leave the course with the foundational understanding of what visual data science is and how it can enhance and transform our approach to working with data and models, as well as practical skills to apply a range of visual data science techniques in their data-intensive, analytical explorations.
Prof. Reza Arghandeh is the leader of the Data Science Group and the director of the Connectivity, Information & Intelligence Lab (Ci2Lab.com) at the Western Norway University of Applied Sciences (HVL), Bergen, Norway. He is also a Lead Data Scientist with StormGeo, an international weather insight company. He was an assistant professor from 2015 to 2018 in the Electrical and Computer Department at Florida State University, USA. Prior to FSU, he was a postdoctoral scholar at the University of California, Berkeley, EECS Dept 2013-2015. He completed his Ph.D. in Electrical Engineering at Virginia Tech. His research interests include spatiotemporal data analysis and computer vision for infrastructure networks. His research has been supported by the U.S. National Science Foundation, the U.S. Department of Energy, the European Space Agency, the European Commission, and the Research Council of Norway.
Abstract: Unlock the potential of data-driven decision-making with our crash course on Applied Causal Inference. Dive into the transformative world of causal inference and gain essential knowledge and practical skills to navigate complex causal relationships within your data. This crash course introduces foundational principles behind causal inference, including interventions, directed acyclic graphs, and structural causal models through a dynamic blend of lectures, real-world case studies, and hands-on exercises. Whether in economics, engineering, healthcare, or social sciences, this course empowers you to ask critical questions and solve complex problems confidently. This crash course focuses on practicality and provides some necessary Python libraries to apply causal inference techniques in your field. Explore causal relationships, identify hidden insights, and make informed decisions that drive meaningful outcomes.
We are happy to announce Professor Iain George Johnston as guest speaker on our statistics theme, at the upcoming Joint CEDAS – NORBIS Summer School 2023.
Iain is a Professor in the Department of Mathematics at UiB. His research group combines mathematical and statistical modelling with bioinformatics and lab work to address questions about evolutionary and cell biology. Current research topics include how cellular power plants in humans, crops, and other organisms evolve and evade damage, and how antimicrobial resistance evolves in dangerous bacteria. You can read more about their work here https://org.uib.no/stochasticbiology/ .
Title: How model selection can turn maths into science
Abstract: Researchers in the business of modelling biology (and more) often fall in love with a particular model of a system. The model is constructed, explored, often tested with unseen data, then published. But this is often misaligned with “science” as empirically understood — the testing of hypotheses with experiment. Any model will fit data to some extent, and admit some theoretical analysis. Why should I believe your model over mine?
On Tuesday and Wednesday, I will discuss the case for a more open relationship with our models. By constructing many different hypothetical descriptions and selecting between them given observations, we can indeed test hypotheses with (numerical) experiments, and engage more directly with open questions about mechanisms and predictions. I’ll discuss some examples of where Bayesian model selection has helped us make biological advances that would be challenging in the lab — including addressing questions about evolution, disease progression, and crop development.
We are delighted that Prof emeritus Arvid Lundervold has agreed to give the opening keynote talk at the Joint CEDAS-NORBIS Summer School 2023.
Title: AI and predictive modelling y ≈ f(X,θ) in medicine and biology
Abstract: This talk aims to inspire and motivate participants, offering a blend of neuroscience, AI, and predictive modelling, demonstrating the power of interdisciplinary thinking in biomedicine. We will discuss the exciting intersection of AI and predictive modelling, and its transformative potential in medicine and biology. We will make use of the predictive modelling framework “y ≈ f(X,θ)” and also draw parallels between the nature of generative AI and biological processes, including how the brain seems to implement language, learning, and memory.
We’ll look at some real-world examples, showcasing how these technologies can revolutionize healthcare and biological research. From predicting disease progression to personalizing treatments, these tools are becoming integral in our quest to understand and improve life and health.
The talk will also address ethical and technological challenges, emphasizing the importance of data privacy, algorithmic transparency, and infrastructure. We will conclude with a forward-looking discussion, highlighting emerging trends and the potential role of early-stage researchers in shaping this exciting field.
Registration is now open for CEDAS and NORBIS members. Emails have been sent to all members. First come, first served. In other words, please register as soon as possible, to be considered for a spot at the event. We will shortly after the opening of the registration look into all registrations and send out a confirmation to those, who we can admit to the school (please note that we have only a limited number of places). Only a confirmed registration enables the actual participation in this summer school.
We are pleased to confirm that Hotel Zander K will host the Joint CEDAS-NORBIS Summer School 2023.