This is a repo to summarize a good course I have taken this semester at Aalto University. The course code is CS-E4820, lectured by Prof. Pekka Martinen. I will update the page after each in-person session of the course summarising the most important parts of the lectures as well the examples solved in session. I also add some extra examples from other sources as well when it is necessary.
Feel free to make a pull request @MLAPM. If you are not familiar with HTML WEB pages, just edit the README.md to add your notes there and I will take care of the html things. Remember to create a pull request when your editing is done.
Acknowledgements: It is worth mentioning that the CSS file for creating this repo is exactly extracted from the great PPC course @ Aalto University and I have used their website as a template for this material. The only difference however is just modifying the CSS file a little bit to add a slideshow functionality to it.
- The examples used in this review are mainly extracted from lecture slides and the reference books:
- Bayesian Reasoning and Machine Learning available at link
- Pattern Recognition and Machine Learning available at link
Review
1. Ingredients of probabilistic modeling:
- Models:
Bayesian networks, Sparse Bayesian linear regression, Gaussian mixture models, latent linear models - Methods for inference:
maximum likelihood, maximum a posteriori (MAP), Laplace approximation, expectation maximization (EM), Variational Bayes (VB), Stochastic variational inference (SVI)
Note:
We don't go deep in MCMC methods because of overlapping with BDA course. The course focuses more on deterministic inference methods. - Ways to select between models
- Marginalization:
getting rid of one of the variables in joint probablity distribution with summation over the other parameter - Independence:
p(b)p(m)=p(b, m)
