CMSC 726: Machine Learning

Course Information

Course Calendar

This calendar is subject to change. All reading is assumed to be done before lecture time. I expect class participation and will avoid rehashing the reading material in lectures. Recommended reading is given in italics.

DateTopicReadingsDue
8/29Introduction
8/31Decision TreesCIML 1
9/3Labor Day NO CLASS
9/5Geometry and nearest neighborsCIML Ch. 2HW00
9/7Practicum I: k-nearest neighbors [notebook] [code]
9/10Perceptrons ICIML 3
9/12Perceptrons IIESL 4.5HW01
9/14Practicum II: perceptrons
9/17Practical issues and evaluation ICIML 4
9/19Practical issues and evaluation IIHW02
9/21Practicum III: Classifier evaluation [notebook] [the map]PA01
9/24Linear models and gradient descentCIML 6
9/26Subgradient descent and support vector machinesHW03
9/28Practicum IV: support vector machines
10/1Probabilistic modeling ICIML 7
10/3Probablistic modeling IIHW04
10/5Practicum V: probabilistic modelingESL 4
10/8Neural networks ICIML 8
10/10Neural networks II
10/12Practicum VI: neural networksPA02
10/15Kernel methods ICIML 9HW05
10/17Kernel methods II
10/19Practicum VII: kernel methods [notebook]
10/22Efficient learningCIML 12
10/24Ensemble methodsCIML 11Midterm
10/26Ensemble methods (cont'd)
10/29Cancelled (weather)
10/31Ensemble methodsSlidesHW06
11/2ClusteringCIML 13
11/5Unsupervised learning (dimensionality reduction)Slides
11/7Expectation maximization ICIML 14HW07
11/9Expectation maximization IIESL 8.5
11/12Semi-supervised learningSSL survey Secs 2-4Slides
11/14Hidden markov modelsSlides Extra ReadingHW08
11/16Hidden markov models
11/19HMMs and Graphical Models
11/21Graphical ModelsSlidesPA03
11/26Graphical ModelsSuggested Reading
11/28Inference in graphical models
11/30Online Learning ISlidesHW09
12/3Online Learning IIIPy Notebook
12/5Bayesian learningSlides Suggested Reading
12/7Deep LearningHinton, et al., 2006 Le, et al., 2012 Bengio, et al., 2007
12/10Structured Output LearningCRF M3N SVM-Struct

PA04,HW10

Written Assignments

These are short written assignments due before class on Wednesdays (unless otherwise noted). Please handin answers in pdf files using the class handin system. Links for the homeworks are not guaranteed to work until the post date.

HomeworkDate postedDue dateHints and solutions
HW00Aug 31Sep 5
HW01Sep 5Sep 12solution
HW02Sep 12Sep 19solution
HW03Sep 19Sep 26solution
HW04Sep 26Oct 3solution
HW05Oct 5Oct 15solution
HW06Oct 26Oct 31
HW07Oct 31Nov 7
HW08Nov 7Nov 14solution
HW09Nov 26Nov 30
HW10Dec 7Dec 10

Midterm

Due on October 24, download here. Use the class handin system to turn it in.

Programming Assignments

Please handin IPython notebook files (and other required source) using the class handin system. Links for the assignments are not guaranteed to work until the post date.

AssignmentDate postedDue date
PA01: Decision Trees [pdf for reference only]Sep 8Sep 21
PA02: Support Vector Machines [pdf for reference only] [data and SVMLib]Sep 28Oct 12
PA03: Ensemble Methods [pdf for reference only] [code and data]Oct 31Nov 21
PA04: Unsupervised Learning[pdf for reference only]Nov 19Dec 10

Project

You will submit a practical project where you apply methods we saw in class (or extensions of them). Your goal is to find a suitable application, decide on methods that are sensible to use for that application and report on your results. We will follow this timetable

WhatDue date
Short writeup describing your application.Nov 9th
Discussion of methods to use and their suitability.Nov 21
Progress report.Dec 7
Final submission.Dec 19

All but the final submission should be posted on Piazza by the due date above.

Resources

Syllabus

Click here for all pertinent administrative information. That page and the class homepage serve as the syllabus for the class.