Advanced ML

  Semi- and unsupervised learning, clustering

  • Task formulation and types of cluster structure
  • Semisupervised learning and classification
  • K-means, DBSCAN and agglomerative hierarchical clustering
  • Co-learnin and co-training
   Slides

  Clustering practice

  • K-means
  • DBSCAN
  • Hierarchical clustering
  • Self-training and co-learning
   ipynb

  Expectation-Minimization algorithm

  • Kullback-Leibler Divergence
  • Motivating example with multiple Gaussians
  • Probabilistic Interpretation of the Principal Component Analysis
   Slides

  EM practice

  • Problem formulation
  • Revise lower-bound definition
  • EM implementation
  • Look at the results
   ipynb

Homework: Unsupervised learning (use JetBrains Edu Plugin)

  Reinforcement learning intro

  • Problem statement and two fundamental problems to solve
  • Examples and cross-entropy method
  • Bellman equations and Temporal difference training
   Slides

  RL intro practice

  • RiverSwim Problem
  • Q-learning implementation
  • Deep Q-Network
   ipynb

Homework: AlphaZero from scratch

  From AlphaGo to AlphaZero

  • TD Gammon
  • Why is Go really hard for computer?
  • Inside AlphaGo: SL and RL policies, MCTS
  • Inside AlphaZero
   Slides

Generative and discriminative models, VAE, GANs

  • Autoencoders
  • Variational Auto Encoders (VAE)
  • Generative adversarial nets
   Slides

Generative models practice

  • Tensorboard
  • Autoencoder: linear, CNN, VAE
  • GAN for MNIST
   ipynb

  CLIP, flows, and diffusion

  • Contrastive Language–Image Pretraining (CLIP)
  • Normalizing flows
  • Diffusion models
   Slides

If you want to use the materials (e.g., figures) in your paper/report and to cite this course, you can do this using the following BibTex:

@misc{avalur2024AdvMLCourse,
title={Advanced ML},
url={https://avalur.github.io/advanced_ml.html},
author={Alexander Avdiushenko},
year={2024},
month={Feb}
}