This series of lectures aims at describing the main problems data scientists and machine/statistical learners have to address (data visualization, dimension reduction, clustering, classification, prediction/regression tasks). For each of these tasks, we will cover several basic strategies that should serve as reference tools at the beginning of any analysis.
The coming lectures are not "theoretical" ones since they do not contain systematic proofs of invovled (but still nice! ) theoretical results.
But their goal is nevertheless to provide guidelines (based on theoretical considerations) for a deeper understanding of the strategies that will be discussed.
For example, the best results are almost never achieved with the default choice of the parameters values. Tuning them carefully depending on the context is what makes the learning strategy work well.
- Enseignant éditeur: Celisse Alain
