Probabilistic Graphical Models

10-708, Spring 2021
School of Computer Science
Carnegie Mellon University


Important Notes

This schedule is tentative and subject to change. Please check back often.

Tentative Schedule

Date Lecture Readings Announcements
Mon, 1-Feb Lecture 1 : Course Introduction
[Slides]

Graphical Models: Representation

Wed, 3-Feb Lecture 2 : Directed Graphical Models: Bayesian Networks
[Slides] [Whiteboard]

Fri, 5-Feb (No Recitation)

Mon, 8-Feb Lecture 3 : Undirected Graphical Models: Markov Random Fields & Conditional Random Fields
[Slides] [Whiteboard] [Poll]

Wed, 10-Feb Lecture 4 : Markov Properties / Factor Graphs
[Slides] [Whiteboard] [Poll]

Fri, 12-Feb Recitation: PyTorch
[Handout]

Graphical Models: Exact Inference and Learning

Mon, 15-Feb Lecture 5 : Exact Marginal/MAP Inference: Variable Elimination & Belief Propagation
[Slides] [Whiteboard] [Poll]

HW1 out

Wed, 17-Feb Lecture 6 : Learning fully observable MRFs and CRFs
[Slides] [Whiteboard] [Poll]

Fri, 19-Feb Recitation: HW1
[Handout]

Mon, 22-Feb Lecture 7 : Neural Potential Functions
[Slides] [Whiteboard] [Poll]

HW1 due

Wed, 24-Feb Lecture 8 : Hybrids of Neural Nets and Graphical Models / Recurrent neural networks (RNNs) / LP, ILP, and MILP
[Slides] [Whiteboard] [Poll]

HW2 out

Fri, 26-Feb Recitation: HW2
[Handout]

Learning for Structured Prediction

Mon, 1-Mar Lecture 9 : MAP Inference: Mixed Integer Linear Programming
[Slides] [Whiteboard] [Poll]

Wed, 3-Mar Lecture 10 : Structured Perceptron / Structured SVM
[Slides] [Whiteboard] [Poll]

Fri, 5-Mar (No Recitation)

Approximate Inference: MCMC

Mon, 8-Mar Lecture 11 : Convolutional neural networks (CNNs)
[Slides] [Whiteboard] [Poll]
  • Monte Carlo Methods. Li (2003). Information Theory, Inference, and Learning Algorithms, Chapter 29 (Section 29.1-29.3).
  • Monte Carlo Methods. Li (2003). Information Theory, Inference, and Learning Algorithms, Chapter 29 (Section 29.4 - 29.5).

Project team matching begins

Wed, 10-Mar Lecture 12 : Complexity of Inference / Monte Carlo Methods
[Slides] [Whiteboard] [Poll]
  • Monte Carlo Methods. Li (2003). Information Theory, Inference, and Learning Algorithms, Chapter 29 (Section 29.6 - 29.10).

HW2 due

HW3 out

Fri, 12-Mar Recitation: HW3 / Project Team Formation Office Hours
[Handout]

Mon, 15-Mar Quiz 1

Wed, 17-Mar Lecture 13 : Markov Chain Monte Carlo: Gibbs Sampling & Metropolis-Hastings
[Slides] [Whiteboard] [Poll]

Fri, 19-Mar No classes: Mid-Semester Break

Mon, 22-Mar Lecture 14 : Markov Chains / Bayesian Inference for Parameter Estimation
[Slides] [Whiteboard] [Poll]

Project team due

Wed, 24-Mar Lecture 15 : Topic Modeling / Particle Filtering
[Slides] [Whiteboard] [Poll]

HW3 due

HW4 out

Fri, 26-Mar Recitation: HW4
[Handout]

Approximate Inference: Variational Methods

Mon, 29-Mar Lecture 16 : Mean Field Variational Inference
[Slides] [Whiteboard] [Poll]

Wed, 31-Mar Lecture 17 : Coordinate Ascent Variational Inference
[Slides] [Whiteboard] [Poll]

Project proposal due

Fri, 2-Apr Lecture 18 : CAVI / Exponential Families / Learning Hidden-state CRFs
[Slides] [Whiteboard] [Poll]

Mon, 5-Apr No classes: Break Day

Wed, 7-Apr Lecture 19 : Learning partially observable graphical models / Variational EM
[Slides] [Whiteboard] [Poll]

HW4 due

HW5 out

Fri, 9-Apr No Recitation

Mon, 12-Apr Lecture 20 : Variational Autoencoders
[Slides] [Whiteboard] [Poll]

Wed, 14-Apr Quiz 2

Recitation: HW5 (7:00pm)

Fri, 16-Apr No classes: Carnival (April 15-17)

Mon, 19-Apr Lecture 21 : Deep Generative Models
[Whiteboard] [Poll]

Advanced Topics

Wed, 21-Apr Lecture 22 : Bayesian Nonparametrics: Dirichlet Process & Indian Buffet Process
[Whiteboard] [Poll]

HW5 due

Fri, 23-Apr Lecture 23 : Bayesian Nonparametrics: Gaussian Process
[Whiteboard] [Poll]

Mon, 26-Apr Lecture 24 : Causality (Part I)
[Whiteboard] [Poll]

Tue, 27-Apr Midway Poster Session

Project midway poster and summary due

Wed, 28-Apr Lecture 25 : Causality (Part II)
[Whiteboard] [Poll]

Fri, 30-Apr Quiz 3

Mon, 3-May Lecture 26 : Reinforcement Learning as Inference
[Whiteboard] [Poll]

Wed, 5-May Lecture 27 : Reinforcement Learning as Inference
[Whiteboard] [Poll]

Fri, 7-May (No Recitation)

Tue, 11-May Project final poster session (5:30pm - 8:30pm) -- details will be announced on Piazza

Project final poster and summary due