Scientists, engineers, and many others routinely collect large and complex data. In general, data visualization can provide insight into the distribution of such data, to confirm the presence of certain patterns or to find patterns not yet known to us. However, if we want to find patterns beyond pairwise relations, it becomes a challenge how to visualize the data to enable this.

Dimensionality reduction (DR) algorithms fill this gap by mapping the data to a 2D or 3D vector space; an embedding of the data. Widely used algorithms include classical methods such as Principal Component Analysis (PCA) and Multi-Dimensional Scaling (MDS), and more recent methods such as t-Stochastic Neighbour Embedding (t-SNE) and Uniform Manifold Approximation and Projection (UMAP). When mapping data to a 2D space, it is not possible to retain all possible structure and each of these methods aims to retain a certain structure from the original data. Hence, there are many different possible approaches, and there are many open questions on the fidelity of various algorithms, which algorithm is best for which application, how to best interpret such embeddings, et cetera.

Hence, a variety of angles on this topic are under active research, such as:

• New approaches to DR for visualization

• Quality measures to assess the fidelity and trustworthiness of embeddings

• Explainability methods and systems supporting interpretation

• Various applications, including cell biology and atmospheric sciences

This workshop is organized in the context of the public PhD defense of ms. Edith Heiter, that takes place August 28th, 9:30 am at Auditorium Vandenhove, Paviljoen Charles Vandenhove, Rozier 1, 9000 Gent. See the public invite.

In this workshop, we aim to bring together

• researchers working on this topic,

• practitioners that use these methods, and

• anyone interested to learn how DR for visualization can be made useful for knowledge discovery and data exploration of complex data.

We are aiming to keep the event free of charge, but we ask you to register in advance to estimate the number of attendees and keep you informed about the event.

Please register here: Registration Form Ghent Workshop on Dimensionality Reduction for Visualization