DTU 02471 ยท Machine Learning for Signal Processing
Problem sets, explained
Each problem has the idea behind it, a small animation of the formula at work, and the solution written out by hand.
Problem set 1
Probability, validation and first signals
Cross-validation, k-NN, k-means, expectation, Bayes, convolution, Fourier series, PCA via Lagrange.
Problem set 2
Filters, estimators and adaptive systems
Filtering as a matrix product, bias and variance, convex norms, AR(1) correlation, Wiener filtering, spectrograms, LMS/NLMS/RLS noise cancelling.