Learning Goals: You should have heard about support vector machines. You should understand Bayesian learning and be able to use Bayes' formula. You should understand the ideas of Gaussian Mixture Models and how it is used in the LDA and QDA algorithms. You should be able to describe how one can learn from unlabeled data and how the EM and k-means algorithms work. You should be able to explain the singular value decomposition (SVD) and how it relates to principal component analysis (PCA).

You find the Lecture 4 slides here (note: a typo on slide 39 has been corrected. The typo is still present in the video...)

Lecture04small.pdf

You find the Lecture 4 video here

or on Youtube here: https://youtu.be/C2nhmFWbNAY

Code used on lecture

svc_iris.ipynb

rbf_parameters.ipynb

bayes.m

pcaMNIST.ipynb (Links to an external site.)

kmeans.ipynb (Links to an external site.)

mixedmodel.m

mixedmodelunsupervised.m

pcacancer.m

The following is a very nice illustration of using SVD to find structure in data (in this case images).

https://colah.github.io/posts/2014-10-Visualizing-MNIST/

For further illustration of use of the SVD I recommend the material about "eigenfaces" available under 1-4 Additional Material - Supervised Learning