Course Description
Machine learning is the study of algorithms that improve their performance at some task with experience. In this course, students will learn how machine learning has led to many innovative realworld applications. The students will also gain an indepth understanding of a broad range of machine learning algorithms from basic to stateoftheart, such as: naïve Bayes, logistic regression, neural networks, clustering, probabilistic graphical models, reinforcement learning and SVMs.
Prerequisites
 50.001 Introduction to Information Systems & Programming and 50.004 Introduction to Algorithms
 Knowledge of programming in Python or Java and a strong foundation in probability and statistics,and optimization (topics such as dynamic programming and bigO notation),and a “B” or better in both 10.004 and 10.009
Learning Objectives
At the end of the term, students will be able to:


 Recognize the characteristics of machine learning that make it useful to realworld problems.
 Explain the basic underlying concepts for supervised discriminative and generative learning.
 Explain the concepts of crossvalidation and regularization, be able to use them for estimation of algorithm parameters.
 Characterize machine learning algorithms as supervised, semisupervised, and unsupervised.
 Have heard of a few machine learning toolboxes.
 Use support vector machines.
 Use regularized regression algorithms.
 Explain the concept behind neural networks for learning nonlinear functions.
 Apply unsupervised algorithms for clustering.
 Explain the foundation of generative models.
 Implement the inference and learning algorithms for the hidden Markov model.
 Explain the learning algorithm for hidden Markov model with latent variables.
 Explain algorithms for learning Bayesian networks.
 Explain reinforcement learning algorithms.

Measurable Outcomes


 List useful realworld applications of machine learning.
 Implement and apply machine learning algorithms.
 Choose appropriate algorithms for a variety of problems.

Topics Covered


 Intro
 Perceptron
 Linear Regression
 Logistic Regression
 Support Vector Machines
 Kernel Methods
 Neural Networks and Deep Learning
 Clustering
 Generative Models
 Mixture Models and Expectation Maximization
 Hidden Markov Model
 Bayesian networks
 Reinforcement Learning

Lecture Schedule


 Cohort session: 3 cohort sessions / week, 5 hours (2+2+1)/ week

Learning Assessment

 Homework, Project, Midterm exam, Final exam