10-701 Fall 2016

Important Notes

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

The first lecture for 10-701 is Wednesday, September 7th, 2016. That is, we start during the second week of classes.

Schedule

Date	Lecture	Topics	Readings	Announcements
Background
Wed, 9/7/16	Lecture 1 (Eric) : Intro to probability, MLE [Slides] [Annotated Slides] [Video]	Course introduction, Basic probability, Maximum likelihood estimate	Probability Tutorial Slides Mitchell. Chap 1,3 Murphy 8.1-3, 8.6 Maximum Likelihood
Supervised Learning
Mon, 9/12/16	Lecture 2 (Eric) : Classification, kNN [Slides] [Annotated Slides] [Video]	Optimal decision using Bayes rule, Types of classifiers, Effect of values of k on kNN classifiers, Probabilistic interpretation of kNN	Fukunuga, Introduction to Statistical Pattern Recognition Ch3 Fukunuga, Introduction to Statistical Pattern Recognition Ch6 Fukunuga, Introduction to Statistical Pattern Recognition Ch7 Liu et al, Clustering Billions of Images with Largest Scale Nearest Neighbor Search Lejsek et al, NV-Tree: Nearest Neighbors at the Billion Scale Lejsek et al, NV-Tree: An Efficient Disk-Based Index for Approximate Search in Very Large High-Dimensional Collections	HW 1 Out
Wed, 9/14/16	Lecture 3 (Matt) : Naive Bayes [Slides] [Video]	Problems with estimating full joints, Advantages of Naive Bayes assumptions, Applications to discrete and continuous cases, Problems with Naive Bayes classifiers	Mitchell, 6.1-6.10
Mon, 9/19/16	Lecture 4 (Matt) : Linear regression [Slides] [Video]	Basic model, Solving linear regression, Error in linear regression, Advanced regression models	Bishop, 3.1
Wed, 9/21/16	Lecture 5 (Matt) : Logistic regression [Slides] [Video]	Logistic regression vs. linear regression, Sigmoid funcion, MLE via gradient ascent, Regularization, Logistic regression for multiple classes	Bishop 4.2-4.3
Mon, 9/26/16	Lecture 6 (Eric) : SVM [Slides] [Annotated Slides] [Video]	Support vector machines, Primal and dual versions of SVM, Duality, KKT conditions	Burges, A Tutorial on Support Vector Machines for Pattern Recognition Fast Training of SVM using Sequential Minimal Optimization	HW2 Out HW1 Due
Wed, 9/28/16	Lecture 7 (Eric) : Kernels & The Kernel Trick [Slides] [Annotated Slides] [Video]	Kernel trick, [SVM Optimization e.g. Sequential Minimal Optimization (SMO)]	Scholkopf and Smola, Learning With Kernels Maximum Entropy Discrimination Markov Networks Maximum Entropy Discrimination
Mon, 10/3/16	Lecture 8 (Eric) : Ensemble learning - Boosting, Random Forests [Slides] [Annotated Slides] [Video]	Combing weak learners, Bagging and random forest, AdaBoost, Algorithm and generalization bounds, Gradient boosting	Freund and Schapire, A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting Schapire et al., Boosting the margin: A new explanation for the effectiveness of voting methods Breiman, Random Forests Friedman, Greedy Function Approximation: A Gradient Boosting Machine Chen, XGBoost, an efficient implementation of gradient boosting	Proposal Due
Theory
Wed, 10/5/16	Lecture 9 (Eric) : Learning theory [Slides] [Annotated Slides] [Video]	Realizable vs agnostic, PAC learning in finite concept class, Sample complexity	Mitchell, Ch. 7 Vapnik, V. (1999) An Overview of Statistical Learning Theory, IEEE transactions on Neural Networks, Vol. 10, pp. 988-99.
Mon, 10/10/16	Lecture 10 (Eric) : VC dimension [Slides] [Annotated Slides] [Video]	Sample complexity for infinite concept classes, VC dimension as a complexity measure, Structural risk minimization	Ch3, An Introduction to Computational Learning Theory, M. Kearns and U. Vazirani	HW3 Out HW2 Due
Wed, 10/12/16	Lecture 11 (Eric) : Evaluating classifiers, Bias-variance decomposition [Slides] [Annotated Slides] [Video]	Bias-variance decomposition, Structural risk minimization, Ways to avoid overfitting	Vapnik V. An Overview of Statistical Learning Theory Andrew Y. Ng, On Feature Selection: Learning with Exponentially many Irrelevant Features as Training Examples
Supervised Learning
Mon, 10/17/16	Lecture 12 (Matt) : Perceptron, Neural networks [Slides] [Video]	Perceptron, Multilayer Perceptron, Backpropagation	Bishop, Ch3 Bengio, Learning Deep Structures for AI Krizhevsky et al,ImageNet Classification with Deep Convolutional Neural Networks
Wed, 10/19/16	Lecture 13 (Matt) : Deep Learning [Slides] [Video]	"Deep" Learning, Convolutional Neural Networks, Layer-wise Pre-training	Learning Learning Deep Generative Models Neural machine translation by jointly learning to align and translate
Unsupervised Learning
Mon, 10/24/16	Lecture 14 (Matt) : PCA and dimension reduction [Slides] [Video]	Principal component analysis, Dimensionality reduction	Murphy, Ch12 Bishop, Ch12	HW4 Out HW3 Due
Wed, 10/26/16	Lecture 15 (Eric) : K-means [Slides] [Video]	Hierarchical clustering, K-means and Gaussian mixture models, Number of clusters	Bishop Ch 9
Mon, 10/31/16	Lecture 16 (Eric) : EM [Slides] [Annotated Slides] [Video]	Expectation Maximization	Blimes, Jeff. A gentle tutorial on the EM algorithm and its application to parameter estimation for Gaussian mixture and hidden Markov Models

Wed, 11/2/16	Midterm
Probabilistic Modeling
Mon, 11/7/16	Lecture 17 (Eric) : Graphical models, Bayes nets [Slides] [Annotated Slides] [Video]	Representation	An Introduction to Graphical Models, M Jordan	Midway Report Due
Wed, 11/9/16	Lecture 18 (Eric) : Inference and learning of graphical models [Slides] [Annotated Slides] [Video]	Learning, Exact Inference	Bishop, 8.4
Mon, 11/14/16	Lecture 19 (Matt) : HMMs and CRFs [Slides] [Video]	Directed vs. undirected, Undirected graphical models, Conditional random fields	Bishop 13.1-13.2 Bishop 8.3-8.4 Sutton & MacCullum, An Introduction to CRF for Relational Learning J. Lafferty, A. McCallum, F. Pereira Conditional Random Fields:Probabilistic Models for Segmenting and Labelling Sequence Data(2001)	HW5 Out HW4 Due
Wed, 11/16/16	Lecture 20 (Matt) : HMMs and CRFs (continued) [Slides] [Video]
Mon, 11/21/16	Lecture 21 (Matt) : Topic modeling and Approximate Inference [Slides] [Video]	Advanced probabilistic modeling, Approximate inference, MCMC	Bishop 11.1- 11.3 Blei(2002) Probabilistic Topic Models Heinrich(2009) Parameter Estimation for Text Analysis Bishop 10.1 - 10.2

Wed, 11/23/16	Lecture (No Class: Thanksgiving) :
Advanced Topics
Mon, 11/28/16	Lecture 22 (Eric) : Distributed ML [Slides] [Video]		E. P. Xing, Q. Ho, P. Xie, W. Dai, Strategies and Principles of Distributed Machine Learning on Big Data, Engineering, Volume:2, pp179 - 95, 2016 E. P. Xing, Q. Ho, W. Dai, J. Kim, J. Wei, S. Lee, X. Zheng, P. Xie, A. Kumar, and Y. Yu, Petuum: A new Platform for Distributed Machine Learning on Big Data, IEEE Transactions on Big Data, Volume:1 Issue:2, pp49 - 67, 2015	HW5 Due
Wed, 11/30/16	Lecture 23 (Matt) : MDPs, Reinforcement learning [Slides] [Video]	Markov decision processes, Value iteration, Policy iteration, Q-Learning	Mitchell Ch. 13 Kaelbling et al. (1996) Silver et al. (2016)
Poster Session
Fri, 12/2/16	Lecture 25 : NSH 3305	2:30 - 5:30 pm
Project Report
Fri, 12/9/16	Project Report Due

Introduction to Machine Learning

Important Notes

Schedule

Background

Supervised Learning

Theory

Supervised Learning

Unsupervised Learning

Probabilistic Modeling

Advanced Topics

Poster Session

Project Report