/umich-eecs545-lectures

This repository contains the lecture materials for EECS 545, a graduate course in Machine Learning, at the University of Michigan, Ann Arbor.

Primary LanguageJupyter Notebook

EECS 545, Winter 2016

This repository contains the lecture materials for EECS 545, a graduate course in Machine Learning, at the University of Michigan, Ann Arbor.

The link above gives a list of all of the available lecture materials, including links to ipython notebooks (via Jupyter's nbviewer), the slideshow view, and PDFs.

Lecture Readings

We will make references to the following textbooks throughout the course. The only required textbook is Bishop, PRML, but the others are very well-written and offer unique perspectives.

Lecture 01: Introduction to Machine Learning

Wednesday, January 6, 2016

No required reading.

Lecture 02: Linear Algebra & Optimization

Monday, January 11, 2016

  • There are lots of places to look online for linear algebra help!
  • Juan Klopper has a nice online review, based on Jupyter notebooks.

Lecture 03: Convex Functions & Probability

Wednesday, January 13, 2016   (Notebook Viewer, PDF File, Slide Viewer)

Required:

  • Bishop, §1.2: Probability Theory
  • Bishop, §2.1-2.3: Binary, Multinomial, and Normal Random Variables

Optional:

  • Murphy, Chapter 2: Probability

Lecture 04: Linear Regression, Part I

Wednesday, January 20, 2016   (Notebook Viewer, PDF File, Slide Viewer)

Required:

  • Bishop, §1.1: Polynomial Curve Fitting Example
  • Bishop, §3.1: Linear Basis Function Models

Optional:

  • Murphy, Chapter 7: Linear Regression

Lecture 05: Linear Regression, Part II

Monday, January 25, 2016   (Notebook Viewer, PDF File, Slide Viewer)

Required:

  • Bishop, §3.2: The Bias-Variance Decomposition
  • Bishop, §3.3: Bayesian Linear Regression

Optional:

  • Murphy, Chapter 7: Linear Regression

Lecture 06: Probabilistic Models & Logistic Regression

Wednesday, January 27, 2016   (Notebook Viewer, PDF File, Slide Viewer)

Required:

  • Bishop, §4.2: Probabilistic Generative Models
  • Bishop, §4.3: Probabilistic Discriminative Models

Optional:

  • Murphy, Chapter 8: Logistic Regression

Lecture 07: Linear Classifiers

Monday, February 1, 2016   (Notebook Viewer, PDF File, Slide Viewer)

Required:

  • Bishop, §4.1: Discriminant Functions

Recommended:

  • Murphy §3.5: Naive Bayes Classifiers
  • Murphy §4.1: Gaussian Models
  • Murphy §4.2: Gaussian Discriminant Analysis

Optional:

Lecture 08: Kernel Methods I, Kernels

Monday, February 8, 2016

Required:

  • Bishop, §6.1: Dual Representation
  • Bishop, §6.2: Constructing Kernels
  • Bishop, §6.3: Radial Basis Function Networks

Optional:

  • Murphy, §14.2: Kernel Functions

Lecture 09: Kernel Methods II, Duality & Kernel Regression

Wednesday, February 10, 2016

Required:

  • Bishop, §6.1: Dual Representation
  • Bishop, §6.3: Radial Basis Function Networks

Optional:

Lecture 10: Kernel Methods III, Support Vector Machines & Gaussians

Monday, February 15, 2016

Required:

  • Bishop, §7.1: Maximum Margin Classifiers
  • Bishop, §2.3.0-2.3.1: Gaussian Distributions

Optional:

Lecture 11: Kernel Methods III, Bayesian Linear Regression & Gaussian Processes

Wednesday, February 17, 2016

Required:

  • Bishop, §3.3: Bayesian Linear Regression
  • Bishop, §6.4: Gaussian Processes

Recommended:

  • Murphy, §7.6.1-7.6.2: Bayesian Linear Regression
  • Murphy, §4.3: Inference in Joinly Gaussian Distributions

Further Reading:

  • Rasmussen & Williams, Gaussian Processes for Machine Learning. (available free online)

Lecture 12: Machine Learning Advice

Monday, February 22, 2016

No required reading.

Lecture 13: Information Theory & Exponential Families

Monday, March 7, 2016

Required:

  • Bishop, §1.6: Information Theory
  • Bishop, §2.4: The Exponential Family

Recommended:

  • Murphy, §2.8: Information Theory
  • Murphy, §9.2: Exponential Families

Further Reading:

Lecture 14: Probabilistic Graphical Models

Wednesday, March 9, 2016

Required:

  • Bishop, §8.1: Bayesian Networks
  • Bishop, §8.2: Conditional Independence
  • Bishop, §8.3: Markov Random Fields

Recommended:

  • Murphy, §10.1: Directed Graphical Models
  • Murphy, §10.2: Examples of Directed Graphical Models

Lecture 15: Latent Variables, d-Separation, and K-Means

Monday, March 14, 2016

Required:

  • Bishop, §8.2: Conditional Independence
  • Bishop, §9.1: K-Means Clustering

Recommended:

  • Murphy, §10.5: Conditional Independence Properties
  • Murphy, §11.1: Latent Variable Models

Lecture 16: Clustering & Expectation Maximization

Wednesday, March 16, 2016

Required:

  • Lecture Notes, "Expectation Maximization" (see Lecture 16 folder)
  • Bishop, §9.2: Mixtures of Gaussians
  • Bishop, §9.3: An Alternative View of EM
  • Bishop, §9.4: The EM Algorithm in General

Recommended:

  • Murphy, §10.3: Inference in Bayesian Networks
  • Murphy, §10.4: Learning in Bayesian Networks
  • Murphy, §11.2: Mixture Models
  • Murphy, §11.3: Parameter Estimation for Mixture Models
  • Murphy, §11.4: The Expectation Maximization Algorithm

Lecture 17: Markov & Hidden Markov Models

Monday, March 21, 2016

Required:

  • Bishop, §13.1: Markov Models
  • Bishop, §13.2: Hidden Markov Models

Recommended:

  • Murphy, §17.2: Markov Models
  • Murphy, §17.3: Hidden Markov Models
  • Murphy, §17.4: Inference in HMMs
  • Murphy, §17.5: Learning for HMMS

Lecture 18: Inference & Applications of Graphical Models

Monday, March 23, 2016

Required:

  • Bishop, §10.1: Variational Inference
  • Bishop, §11.2: Markov Chain Monte Carlo

Recommended:

  • Murphy, §19.1-4: Markov Random Fields
  • Murphy, §21.2: Variational Inference
  • Murphy, §21.3: The Mean Field Method
  • Murphy, §23.1-4: Monte Carlo Inference
  • Murphy, §24.1-3: Markov Chain Monte Carlo
  • Murphy, §27.3: Latent Dirichlet Allocation

Lecture 19: Principal Components Analysis & ICA

Monday, March 28, 2016

Required:

  • Bishop, §12.1: Principal Components Analysis
  • Bishop, §12.2: Probabilistic PCA
  • Bishop, §12.3: Kernel PCA
  • Bishop, §12.4: Nonlinear Latent Variable Models

Recommended:

  • Murphy, §12.2: Principal Components Analysis
  • Murphy, §12.4: PCA for Categorical Data
  • Murphy, §12.6: Independent Component Analysis