[Webinar] Dimensionality reduction, some intuitions from an unusual angle

Dimensionality reduction is one of those tools that feels intuitive until someone asks you a slightly unexpected question about it. This webinar starts from scratch, building geometric intuition that requires no prior knowledge of the subject. But even if you have used PCA for years, the angle we take may offer a new way of seeing something familiar.

From there we visit several topics that rarely make it into standard tutorials: what it really means to reconstruct data after projecting it, how preprocessing choices silently reshape your components, and how to choose the number of components in a principled way, whether you are working in a supervised or an unsupervised setting.

We close with a practical demonstration on text data, tying the concepts together on a concrete problem.