A Practical Guide to Dimensionality Reduction
Dimensionality reduction is an important step in preparing data for both supervised and unsupervised learning. It supports data exploration by helping to reveal patterns in high-dimensional datasets while reducing complexity. This course introduces three main approaches to dimensionality reduction: feature transformation, feature aggregation, and feature selection.
Participants will explore the key methods within each approach and compare their advantages and limitations. Through practical exercises in Jupyter notebooks using a real-world dataset, participants will gain hands-on experience applying these methods and develop a deeper understanding of when and how to use them effectively.
Learning Goals
Part 1: General introduction and feature transformation methods
-
General introduction to dimensionality reduction
-
Theory and practical application of classical feature transformation methods
-
Theory and practical application of autoencoders for feature transformation
Part 2: Further unsupervised and supervised methods
-
Theory and practical application of feature aggregation approaches
-
Theory and practical application of feature selection methods
-
Stability optimization in feature selection
Prerequisites
Basic knowledge of Python. Basic understanding of machine learning models. Google account is recommended.
Target Group
This course is open to researchers of all career stages, or anyone interested in learning about the subject.
This course is open to individuals affiliated with Helmholtz or a HIDA Partner only.