Content and Methodology
Short Description of Methodology
The Winter School will provide a study of several emerging topics of high relevance within advanced deep learning, from a basic understanding of the techniques to the latest state-of-the-art developments in the field. In addition, participants will be exposed to the latest advances and applications in deep learning through oral presentations and poster presentations in the main conference program.
The course will consist of 5 full days of the NLDL conference, including tutorials, keynote sessions, oral presentations, poster presentations, and practical components.
This 5-day course is built upon tutorials on specific topics in deep learning. The course further encompasses, among other things, keynote talks as well as special sessions on industry and diversity in AI as part of the NLDL conference program. The course content will be made available digitally in the form of recorded talks.
Tutorial 1: Non-Euclidean Deep Learning
Within deep learning, Euclidean geometry is the default basis for deep neural networks. However, the naive assumption that such a topology is optimal for all data types and tasks does not necessarily hold. This tutorial will provide an introduction to non-Euclidean structures such as hyperbolic and hyperspherical spaces and their usefulness in deep learning. By the end of the tutorial, attendees will have a comprehensive understanding of hyperbolic and hyperspherical spaces in the context of deep learning.
The tutorial will be presented by Aiden Durrant from the University of Aberdeen.
Tutorial 2: Large Language Models-Based Copilots and Artificial Intelligence Agents
This tutorial presents cutting-edge research on the use of large language model copilots for automatic but human-guided processing of complex data sources. Participants will learn how large data-driven tools based on large language models can provide robust and context-aware data processing for non-technical domain experts. By the end of the tutorial, attendees will have gained knowledge about the latest advancements in large language model-based copilots and hands-on experience with using such tools.
The tutorial will be presented by Mihaela van der Schaar from the University of Cambridge and Anders Boyd from Amsterdam University Medical Centers.
Tutorial 3: Multimodal Learning
Fusing information from multiple modalities is a fundamental challenge in machine learning. Recent works have achieved impressive performance on a wide range of tasks involving modalities from various sources such as images, time series, and text. Recent multimodal approaches heavily rely on deep neural networks to process various data types and fuse them into a common representation that contains complementary information from the different sources. Performing this fusion in a reliable and precise manner is key to achieving good performance.
This course will provide both a fundamental understanding of the techniques and methods used in multimodal learning and the most recent innovations within the field. The course will cover multimodal learning in the context of important data domains such as image, text, and time series data, giving students a deeper understanding of the theory underpinning current multimodal methods.
The tutorial will be presented by Kristoffer Wickstrøm and Michael Kampffmeyer from UiT The Arctic University of Norway.
