This course consists of 4 whole days (2x2).
Day 1: Data Visualisation Principles and Techniques
Instructor: Prof. Katrien Verbert (KU Leuven)
This day introduces the core principles and techniques of data visualisation. Participants will explore how human perception affects data interpretation, learn effective methods for data representation, and understand the role of interaction in enhancing visualisations. The day will conclude with a practical Tableau tutorial, allowing participants to apply these concepts hands-on.
Day 2: Visualisation for Data Science & AI: Data Preparation & Model Quality Evaluation
Instructor: Prof. Jefrey Lijffijt (UGent)
This day focuses on the integration of visualisation in the machine learning development pipeline. Participants will learn about the role of visualisation in data preparation and model quality evaluation. Key topics include:
- Exploratory Data Analysis (EDA): Using various plots such as histograms, density/KDE plots, raincloud plots, empirical CDF plots, correlation heatmaps, and pair plots.
- Dimensionality Reduction (DR) Methods & Representation Learning: Understanding captured information, the elbow method, and silhouette analysis.
- Handling Missing Values: Creating missing value plots.
- Outlier Detection: Techniques like local-outlier factor and tools like Cleanlab.
- Contrasting Data & Distributions: Effective methods for comparison.
- Model Evaluation: Using AUC-ROC, PR curves, hyper-parameter sensitivity, calibration, and monitoring tools like Weights & Biases and TensorBoard.
- Bias and Fairness: Examples and toolboxes, including the visual auditor.
Day 3: Human-Centered Explainable AI
Instructor: Prof. Katrien Verbert (KU Leuven)
This day is dedicated to the human-centred aspects of explainable AI. Participants will delve into:
- Human-Centered Explainable AI: Understanding the importance of making AI systems interpretable and user-friendly.
- Interaction Design: Techniques for designing effective user interactions with AI systems.
- Cognition: Gaining insights into how users perceive and understand AI outputs.
- Human-Centered Design Methods: Approaches to designing AI systems with the user in mind.
- Evaluation Methods: Techniques for assessing the effectiveness and usability of AI systems.
- Hands-On Activity: Practical exercises to apply the concepts learned throughout the day.
Day 4: AI for Visualisation: Dimensionality Reduction and Automation
Instructor: Prof. Jefrey Lijffijt (UGent)
The final day explores the application of AI in data visualisation, focusing on dimensionality reduction and automation. Key topics include:
- The why/what/how of visualization & what parts of the data visualization pipeline can we automate? Discussion of survey papers & practical examples: what to look at and how to look at it (approaches using traditional ML/optimization as well as GenAI).
- Dimensionality Reduction (DR): Linear and non-linear DR methods (PCA, LDA, autoencoders, MDS, and t-SNE), why there exist so many methods, categorizing them and learning when to use what type of method.
- Representation learning as an intermediate step, with practical examples of graph data.
The lecturers
Katrien Verbert is a professor at the Augment research group of KU Leuven. She obtained a doctoral degree in Computer Science in 2008 at KU Leuven, Belgium. She was a postdoctoral researcher of the Research Foundation – Flanders (FWO) at KU Leuven. She was an Assistant Professor at TU Eindhoven, the Netherlands (2013 –2014) and Vrije Universiteit Brussel, Belgium (2014 – 2015). Her research interests include visualization techniques, recommender systems, explainable AI, and visual analytics. She has been involved in several European and Flemish projects on these topics, including the EU ROLE, STELLAR, STELA, ABLE, LALA, PERSFO, Smart Tags and BigDataGrapes projects.
Jefrey Lijffijt is a professor of Data Science, Knowledge Discovery, and Visual Analytics at Ghent University, where he teaches the master-level course ‘Data Visualisation for and with AI’ at the Faculty of Engineering and Architecture. He has a background in algorithms, statistics, machine learning, and data visualization. He likes making tools that help others to better understand and utilize data.
