04 Methodology 2: Data cleaning, Principal Component Analysis (MLVU2018) @riskone1

Uploaded February 2018 | Updated September 2026, 6 days ago

Errata: I mix up the eigendecomposition and singular value decomposition in this one. I recommend watching the 2019 version instead.

Lecture 4 in the Machine Lecture course at the VU University Amsterdam. Lecturer: Peter Bloem. See the PDF for image credits.

Today we discuss how to treat data before running it though a machine learning algorithm. How to deal with missing data, how to deal with outliers how to define your features, and how to normalize your data. We also discuss principal component analysis, a very powerful data normalization method.

Corrections:
- On slide 31, the formula for normalization is incorrect. x_min should be subtracted, not added.

Slides: dropbox.com/s/3okpymh4adspovg/22.Methodology2.annotated.pdf?dl=0

04 Methodology 2: Data cleaning, Principal Component Analysis (MLVU2018)

MLVU 1.3 Other abstract tasks: regression, clustering, density estimation

MLVU 9.2: Generative adversarial networks (GANs)

1 Introduction to Machine Learning (MLVU2020)

8 Probability 2: Maximum Likelihood, Gaussian Mixture Models and Expectation Maximization (MLVU2019)

6 Linear Models 2: Neural Networks, Backpropagation, SVMs and Kernel methods (MLVU2019)

11 Sequential Data: Markov Models, Word Embeddings and LSTMs