Orateur
Martin Skrodzki
Description
The field of explorative data analysis provides methods to investigate large, potentially high-dimensional, data sets. Such exploration is best done visually, to engage the human in the loop with all pattern recognition built into our visual cortex. In my talk, I will give a brief introduction to dimensionality reduction for this purpose, especially about the t-SNE method. I will then show how embedding into non-Euclidean spaces can provide embeddings that help with visual inspection and how to compute such embeddings quickly.