Practical machine learning: methods and algorithmics

• News
• Tentative Course Schedule
• Resources
• Course Information
• Assignments
• Project

News

• (3/19) Here are contest results: txt. Please claim your prizes during today's office hours at 1p, or email Hector
• (3/17) The contest for HW2 is taking a little long to run, so contest results will be posted here on Friday at noon. Visit office hours Friday at 1p to claim your prize, if applicable.
• (3/17) Lecture slides on clustering pdf
• (3/15) Homework 3 has been updated
• (3/15) Lecture slides on Ensemble methods pdf
• (3/12) Homework 3 is posted here
• (3/11) Homework 3 will be posted later today...
• (3/10) For today, we'll use Andrew Moore's (at CMU) slides on Support Vector Machines pdf
• (3/4) Added a little example of using cross-validation for ridge and lasso model selection and assessment R code
• (3/2) Notes for lecture 7 are now posted pdf
• (3/1) Notes for lecture 6 have been posted pdf
• (2/27) Homework 2 is now posted. Due Monday March 8.
• (2/24) Notes for lecture 5 have been updated
• (2/22) Notes for lecture 5 have been posted pdf R code
• (2/18) See more information about the project here
• (2/18) I have added official office hours. See here
• (2/17) The following R code shows how to run a few classification methods on a dataset of prostate patients
• (2/15) Homework 1 has been posted. Due February 22
• (2/15) Now that we've dug out, let's continue with class... For today we will finish linear methods for classification and do tree-based methods.
• (2/3) Check the resources section to get data and code for lecture note scripts
• (2/1) Notes for lecture 3 have been posted. Download pdf here. (R code)
• (1/31) Notes for lecture 2 have been updated. (pdf) (R code)
• (1/27) Notes for lecture 2 are posted. Download pdf here. (R code)
• (1/25) Notes for lecture 1 are posted. Download pdf here. (R code)
• (1/25) Notes on background material (lecture
  0. are posted. Download pdf here.
• (1/25) This section will have notifications to the class. A notification will be made here when lecture notes and assignments are ready for download.

Tentative Course Schedule

0. Background (pdf)
1. Introduction to regression and prediction (1/25) (pdf) (R code)
2. Linear methods for regression (1/27,2/1) (pdf) (R code)
3. Linear methods for classification (2/3,2/15) (pdf) (R code)
4. Tree-based methods for regression and classification (2/15) (pdf) (R code)
5. Model selection and assessment (2/22,2/24) (pdf) (R code)
6. Methods for high-dimensional problems (3/1) (pdf)
7. Smoothing (3/3, 3/8) (pdf)
   
8. Support vector machines (3/10) (pdf)
9. Ensemble methods (3/15) (pdf)
10. Unsupervised methods (3/17) (pdf)

Resources

• Data

  • CD4 data (text file)

  • Datasets from Hastie, Tibshirani and Friedman's book are available as an R package. Try the following in R to get a list of datasets:

    > install.packages("ElemStatLearn")
    > library(help=ElemStatLearn)
    

  • There are number of datasets available in the mlbench R package as well:

    > install.packages("mlbench")
    > library(help=mlbench)
    

  • Olive oils data (text file)

• Code

  • Univariate knn regression (R code)
  • Rosenblatt's perceptron algorithm (R code)

• Other

  • This page is a good place to see some of the methods we are using in class in action
  • The Machine Learning CRAN Task View is also a good place to check out

Course Information

• Instructors

  • Héctor Corrada Bravo
    Bloomberg School of Public Health E3140
    
    Office hours: M 3-4, Th 12-1, F 1-2

  • Rafael A. Irizarry
    Bloomberg School of Public Health E3619
    

• TA: Hao Wu
  Bloomberg School of Public Health E3037
  

• Meeting times

  Monday and Wednesday 1:30pm-2:50pm
  JHU Bloomberg School of Public Health
  615 N. Wolfe Street
  Room W4019

• Evaluation

  There will be three assignments including programming, data analysis and pen-and-paper problems. There will also be a graded open project.

• Materials

  There is no required textbook. Lecture notes and slides will be made available before each lecture. Recommended books for reference are

  • The Elements of Statistical Learning, T. Hastie, R. Tibshirani and J. Friedman.
  • Pattern Recognition and Machine Learning, C. Bishop.
  • Machine Learning, T. Mitchell.
  • Pattern Classification, R. Duda, P. Hart and D. Stork
  • Pattern Recognition and Neural Networks, B. Ripley

  It is required that programming assignments are done with R. For the project, programs can be written in any other language. However, instructors and the TA must be able to compile and run them in the Biostats department computing environment.

• Description

  Machine Learning methods have found their way into the modern data analyst's toolbox. This course introduces popular methods with an emphasis on their practical usage for data analysis. Some basic coverage of their statistical and computational theoretical underpinnings acquaints students with methods to evaluate statistical machine learning models defined in terms of algorithms or function approximations.

  Topics covered include: regression and prediction, tree-based methods, overview of supervised learning theory, support vector machines, kernel methods, ensemble methods, clustering, visualization of large datasets and graphical models.

  Examples of method applications covered include cancer prognosis from microarray data, visualization and analysis of social network data, and graphical models for clinical decision-making.

  Prerequisites: Linear algebra and statistical principles at a 651-654 level. Proficiency in R, or programming language with suitable numerical capabilites (e.g. matlab, python). Listed in catalog as 160.644.

Assignments

1. Homework 1 can be found here
2. Homework 2 can be found here
3. Homework 3 can be found here

Project

There is an open project. Details here