stimaliu/advmlcourse

Course repo for Advanced Machine Learning Course at Linköping University

advmlcourse's Issues

Show random fields

Show pictures of GP realizations for bivariate inputs. Additive and non-additive.

Lab 3 - Gaussian processes

• Question 1. Ask the students to write the kernel functions so that they are compatible with kernlab. Smoother transition to question 2.
• Question 2. Ask the students to add 95% bands also on the model with day and the locally periodic kernel. Compare the models.
• Question 3 - classification. This one was too easy. Add at least that also logistic regression is fitted, plot the decision contours and compute accuracy. Think of some other analysis to do here. Too thin.

Gaussian process - show correlation function

When presenting kernels, show also the correlation function. Plot of k(|x-x'|) vs |x-x'|.

GP lectures

• Present the Cholesky for GP regression
• Show the posterior also for a general prior mean function. Re-analyze the canadian data with a better mean. Show one can implement a general mean via the kernel (see Bishops book).