10-708 PGM | Schedule

Date	Lecture	Readings	Logistics
Module 1: Introduction, Representation, and Exact Inference
1/14	Lecture #1 (Eric): Introduction to GM [ slides (annotated) \| video \| notes ]	Jordan, Graphical Models E. Airoldi, Getting Started in Probabilistic Graphical Models
1/16	Lecture #2 (Eric): Representation: Directed GMs (BNs) [ slides (annotated) \| video \| notes ]	Jordan Textbook, Ch. 2 (Sec. 2.1) Koller and Friedman Textbook, Ch. 3
1/21	No class (MLK day)
1/23	Lecture #3 (Eric): Representation: Undirected GMs (MRFs) [ slides (annotated) \| video \| notes ]	Jordan Textbook, Ch. 2 (Sec. 2.2 - end) Koller and Friedman Textbook, Ch. 4 A. Fischer and C. Igel, An Introducton to Restricted Boltzmann Machines
1/28	Lecture #4 (Eric): Exact inference - Elimination - Message passing - Sum product algorithm [ slides (annotated) \| video \| notes ]	Jordan Textbook, Ch. 3 and 4 Koller and Friedman Textbook, Ch. 9, Ch. 10 Tom Minka, Divergence measures and message passing	HW1 out (Mon, 1/28)
1/30	Lecture #5 (skipped): Parameter learning in fully observable Bayesian Networks - Generalized Linear Models (GLIMs) - Maximum Likelihood Estimation (MLE) - Markov Models [ slides \| video \| notes ]	Jordan Textbook, Ch. 8, 9 (Sec. 9.1 - 9.2) Koller and Friedman Textbook, Ch. 17 (Sections 17.1 - 17.4) Geiger and Heckerman, Parameter priors for directed acyclic graphical models and the characterization of several probability distributions Geiger and Heckerman, A characterization of the Dirichlet distribution through global and local parameter independence
2/4	Lecture #6 (Maruan): Parameter Learning of partially observed BN - Mixture models - Hidden Markov Models (HMMs) - The EM algorithm [ slides (annotated) \| video \| notes ]	Jordan Textbook, Ch. 11 Koller and Friedman Textbook, Ch. 19.1-19.4 Borman, The EM algorithm (A short tutorial) Some interesting aspects of the EM algorithm (Neal and Hinton)
2/6	Lecture #7 (Eric): Maximum likelihood learning of undirected GM [ slides (annotated) \| video \| notes ]	Jordan Textbook, Ch. 9 (Sec. 9.3-9.5), Ch. 20 Friedman, Hastie, Tibshirani. Sparse inverse covariance estimation with the graphical lasso. Lafferty et al., Conditional Random Fields - Probabilistic Models for Segmenting and Labeling Sequence Data.
2/11	Lecture #8 (guest lecture, Kun Zhang @ Department of Philosophy): Causal discovery and inference [ slides \| video \| notes ]	J. Pearl, M. Glymour, and N. P. Jewell. Causal Inference in Statistics A Primer. Concepts related to causality and identification of causal effects. P. Spirtes, C. Glymour, and R. Scheines. Causation, Prediction, and Search. Pages 80-89 and 125-123-136. Traditional methods for causal discovery, including the PC and FCI algorithms. K. Zhang, B. Schöklopf, P. Spirtes, and C. Glymour. Learning causality and causality-related learning. Brief review of recent methods for causal discovery.
2/13	Lecture #9 (Eric): Modeling networks - Gaussian graphical models - Ising models [ slides (annotated) \| video \| notes ]	Meinshausen and Buhlmann, High-Dimensional Graphs and Variable Selection with the Lasso. Kolar et al., Estimating Time-Varying Networks. Dempster, Covariance Selection.	HW1 due (Wed, 2/13)
2/18	Lecture #10 (Eric): Sequential models - Discrete Hidden State (HMM vs. CRF) - Continuous Hidden State (Kalman Filter) [ slides (annotated) \| video \| notes ]	Jordan Textbook, Ch. 14, Ch. 15 Ng, Lecture notes on factor analysis. Welch and Bishop, An Introduction to the Kalman Filter. Conditional Random Fields - An Introduction.
Module 2: Approximate Inference
2/20	Lecture #11 (Eric): Approximate Inference: Mean Field (MF) and loopy Belief Propagation (BP) approximations [ slides (annotated) \| video \| notes ]	Yedidia et al., Generalized Belief Propagation. Xing et al., A Generalized Mean Field Algorithm for Variational Inference in Exponential Families. Variational inference tutorial (NIPS 2016)	HW2 out (Fri, 2/22) Project proposal due (Fri, 2/22)
2/25	Lecture #12 (Eric): Theory of Variational Inference: Inner and Outer Approximations [ slides (annotated) \| video \| notes ]	Wainwright and Jordan, Variational Inference in Graphical Models - The View from the Marginal Polytope. Wainwright and Jordan, Graphical Models, Exponential Families, and Variational Inference (Sec. 3 and 4).
2/27	Lecture #13 (Eric): Approximate Inference: Monte Carlo and Sequential Monte Carlo methods [ slides (annotated) \| video \| notes ]	Jordan Textbook, Ch. 21 David MacKay's Textbook, Ch. 29 (Sec. 29.1-29.3).
3/4	Lecture #14 (Eric): Markov Chain Monte Carlo - Metropolis-Hastings - Hamiltonian Monte Carlo - Langevin Dynamics [ slides (annotated) \| video \| notes ]	David MacKay's Textbook, Ch. 29 (Sec. 29.4-29.10). Neal, MCMC using Hamiltonian dynamics. Patterson and Teh, Stochastic Gradient Riemannian Langevin Dynamics on the Probability Simplex.
Module 3: Deep Learning & Generative Models
3/6	Lecture #15 (Eric): Statistical and Algorithmic Foundations of Deep Learning - Insight into DL - Connections to GM [ slides (annotated) \| video \| notes ]	Deep learning book, Ch. 6.2-5, 20.3-4. Salakhutdinov and Hinton, Deep Boltzmann machines. Belanger and McCallum, Structured prediction energy networks. Ranganath et al., Deep exponential families.	HW2 due (Mon, 3/11)
3/11	No classes (Spring break)
3/13	No classes (Spring break)
3/18	Lecture #16 (guest lecture, Zhiting Hu): Building blocks of DL - RNN and LSTM - CNN, Transformers - Attention mechanisms - (Case studies in NLP) [ slides \| video \| notes ]	Pascanu, Mikolov, Bengio, On the difficulty of training recurrent neural networks. Vasvani et al., Attention is all you need. Devlin et al., BERT - Pre-training of Deep Bidirectional Transformers for Language Understanding.	HW3 out (Mon, 3/18)
3/20	Lecture #17 (Eric): Deep generative models (part 1): Overview of the theoretical basis and connections of deep generative models - Wake sleep algorithm - Variational autoencoders - Generative adversarial networks - A unified view of DGM [ slides \| video \| notes ]	Deep learning book, Ch. 20.9-10 Variational Autoencoders, Kingma and Welling, 2014. Generative Adversarial Nets, Goodfellow et al., 2014. Sanjeev Arora's blog post on GANs.
3/25	Lecture #18 (guest lecture, Zhiting Hu): Deep generative models (part 2) - GANs and their variations - Normalizing Flows - Integrating domain knowledge in DL [ slides \| video \| notes ]	Arjovsky, Bottou, Towards principled methods for training generative adversarial networks. Kingma, Dhariwal, Glow - Generative Flow with Invertible 1x1 Convolutions. Hu et al., Harnessing Deep Neural Networks with Logic Rules. Hu et al., Deep Generative Models with Learnable Knowledge Constraints.
3/27	Lecture #19 (guest lecture, Zhiting Hu): Case Study: Text Generation - The encoder-decoder framework - Machine translation as conditional generation - Unifying MLE and RL objectives for text generation [ slides \| video \| notes ]	Ranzato et al., Sequence Level Training with Recurrent Neural Networks. Hu et al., Toward Controlled Generation of Text. Tan, Hu et al., Connecting the Dots Between MLE and RL for Sequence Generation.	Midway report due (Fri, 3/29)
Module 4: Reinforcement Learning & Control Through Inference in GM
4/1	Lecture #20 (Maruan): Sequential decision making (part 1): The framework - Brief introduction to reinforcement learning (RL) - Connections to GM: RL and control as inference - Control via Variational Inference [ slides (annotated) \| video \| notes ]	Sutton, Barto, Reinforcement Learning, An Introduction, Ch. 3, 4. Lilian Weng, A peek into RL (blog). Levine, Reinforcement Learning and Control as Probabilistic Inference, Sec. 1-3. Ziebart, Modeling Purposeful Adaptive Behavior with the Principle of Maximum Causal Entropy, Ch. 5.1-2, 6.1-2. Todorov, General duality between optimal control and estimation. Koller and Friedman Textbook, Ch. 20.3.
4/3	Lecture #21 (Maruan): Sequential decision making (part 2): The algorithms - Intro to RL algorithms: policy gradients and Q-learning - Max-entropy policy gradient - Soft Q-learning [ slides (annotated) \| video \| notes ]	Sutton, Barto, Reinforcement Learning, An Introduction, Ch. 13. Levine, Reinforcement Learning and Control as Probabilistic Inference, Sec. 4. Haarnoja et al., Reinforcement Learning with Deep Energy-Based Policies. Haarnoja et al., Soft Actor-Critic.	HW3 due (Wed, 4/3)
Module 5: Nonparametric methods
4/8	Lecture #22 (Eric): Bayesian non-parameterics - Dirichlet Process (DP) - Indian Buffet Process (IBP) [ slides (annotated) \| video \| notes ]	Teh, Dirichlet Process. Griffiths and Ghahramani, The Indian Buffet Process.	HW4 out (Mon, 4/8)
4/10	Lecture #23 (Maruan): Bayesian non-parameterics (continued) - Inference in Dirichlet Process (DP) - Hierarchical Dirichlet Process (HDP) - Indian Buffet Process (IBP) [ slides \| video \| notes ]	Notes on Dirichlet Processes. Ishwaran, James, Gibbs Sampling Methods for Stick-Breaking Priors. Teh, Jordan, Beal, Blei, Hierarchical Dirichlet Process.
4/15	Lecture #24 (Eric): Integrative Paradigms of GM: Regularized Bayesian Methods [ slides (annotated) \| video \| notes ]	Ganchev, Graça, Gillenwater, Taskar. Posterior Regularization for Structured Latent Variable Models. Zhu, Chen, Xing. Bayesian Inference with Posterior Regularization and Applications to Infinite Latent SVMs.
4/17	Lecture #25 (Eric): Elements of Spectral & Kernel GMs [ slides (annotated) \| video \| notes ]	Song. Learning via Hilbert space embedding of distributions, Sec. 2.1, 2.2, 3.1, 3.2. Hsu, Kakade, Zhang. A Spectral Algorithm for Learning Hidden Markov Models. Song, Liu, Parikh, Xing. Nonparametric latent tree graphical models. Parikh, Song, Xing. A Spectral Algorithm for Latent Tree Graphical Models.
4/22	Lecture #26 (Maruan): Gaussian processes (GPs) and elements of meta-learning - Gaussian Processes (GPs) and kernel functions - (Deep) kernel learning and approximations - Neural Processes (NPs) as an approximation to GPs - Elements of meta-learning [ slides \| video \| notes ]	Rasmussen, Williams. Gaussian Processes for Machine Learning, Ch. 2.2-2.4, 4.1-4.2. Görtler, Kehlbeck, Deussen. A Visual Exploration of Gaussian Processes. Wilson, Hu, Salakhutdinov, Xing. Deep Kernel Learning. Garnelo et al. Conditional Neural Processes.
Module 6: Modular and scalable algorithms and systems
4/24	Lecture #27 (guest lecture, Qirong Ho): Scalable algorithms and systems for learning, inference, and prediction [ slides \| video \| notes ]	Bradley et al. (2011). Parallel Coordinate Descent for L1-Regularized Loss Minimization. Yuan et al. (2015). LightLDA - Big Topic Models on Modest Computer Clusters. Ho et al. (2013). More effective distributed ML via a stale synchronous parallel parameter server. Kim et al. (2016). STRADS - A Distributed Framework for Scheduled Model Parallel Machine Learning.	HW4 due (Mon, 4/24)
4/29	Lecture #28 (Eric): A civil engineering perspective on AI [ slides (annotated) \| video \| notes ]	Xing et al. Petuum - A New Platform for Distributed Machine Learning on Big Data. Xing et al. Strategies and Principles of Distributed Machine Learning on Big Data.
4/30	Project presentations (NSH attrium, 2:30-5:30 pm)