Machine Learning

This award-winning introductory Machine Learning lecture teaches the foundations of and concepts behind a wide range of common machine learning models. It uses a combination of engaging lectures, challenging mathematical exercises, practically-oriented programming tasks, and insightful tutorials. The lecture was awarded with the TeachInf 2020 award.

The Machine Learning lecture for WS21/22 will again be held online. We will upload videos of lectures and tutorials, and provide pointers to other reference materials. Additionally, we will offer slots for online, live Q&A sessions (every Wednesday from 12 to 2pm).

Important: The first session, in which we will discuss organizational topics, will be held live on Wednesday October 20th at 1pm (not 12pm!).
A link to the zoom call is available on the Piazza forum and Moodle.

Please do not send any questions about organizational matters via e-mail. Use the Q&A session and, after that, the Piazza forum.
If you have problems accessing the Moodle course, contact jan.schuchardt [at] in.tum.de .

Tentative list of topics

  • Introduction
    • What is machine learning?
    • Typical tasks in ML
  • k-Nearest neighbors
    • kNN for classification and regression
    • Distance functions
    • Curse of dimensionality
  • Decision trees
    • Constructing & pruning decision trees
    • Basics of information theory
  • Probabilistic inference
    • Parameter estimation
    • Maximum likelihood principle
    • Maximum a posteriori
    • Full Bayesian approach
  • Linear regression
    • Linear basis function models
    • Overfitting
    • Bias-variance tradeoff
    • Model selection
    • Regularization
  • Linear classification
    • Perceptron algorithm
    • Generative / discriminative models for classification
    • Linear discriminant analysis
    • Logistic regression
  • Optimization
    • Gradient-based methods
    • Convex optimization
    • Stochastic gradient descent
  • Deep learning
    • Feedforward neural networks
    • Backpropagation
    • Structured data: CNNs, RNNs
    • Training strategies
    • Frameworks
    • Advanced architectures
  • Support vector machines
    • Maximum margin classification
    • Soft-margin SVM
  • Kernel methods
    • Kernel trick
    • Kernelized linear regression
  • Dimensionality reduction
    • Principal component analysis
    • Singular value decomposition
    • Probabilistic PCA
    • Matrix factorization
    • Autoencoders
  • Clustering
    • k-means
    • Gaussian mixture models
    • EM algorithm

Literature

  • Pattern Recognition and Machine Learning. Christopher Bishop. Springer-Verlag New York. 2006.
  • Machine Learning: A Probabilistic Perspective. Kevin Murphy. MIT Press. 2012

Prerequisites

  • Good understanding of Linear Algebra, Analysis, Probability and Statistics.
  • Programming experience (preferably in Python).