Lecture: Tibshirani's notes from Chapter 4 and Chapter 4 of the text

Homework: Homework03.pdf

R-code:

     Here's the code for the Wed class:  ZipCode.R

     Code for IRLS two-class logistic regression LogisticRegression.R

     Examples using glmnet including South African Heart Study used in ESL and digits 1-4 of the zip code data glmnetExamples.R

Example Data:

     Graphical and text output for in-class example:

     text files corresponding to the two plots above:

          Err_4var_target.txt

          Err_7var_target.txt    

To review matrix derivatives, check out page 8 of Stanford Professor Andrew Ng's notes:

http://www.stanford.edu/class/cs229/notes/cs229-notes1.pdf

http://stanford.edu/class/cs229/section/cs229-linalg.pdf 

Here's a link to paper by Friedman, Hastie, Tibshirani paper Regularization Paths for Generalized Linear Models via Coordinate Descent,  (Journal of Statistical Software, Feb 2010).  This paper describes the algorithm used in the glmnet R-package.  http://www.jstatsoft.org/v33/i01/

Here's the plot cv output from glmnet classifying characters 1-4 from the handwritten zip code data set.  damn glmnet is fast!

Recorded Lecture 3:

Part 1: https://datamining.webex.com/datamining/ldr.php?AT=pb&SP=MC&rID=107355092&rKey=c4366fcaae331ce6

Part 2: https://datamining.webex.com/datamining/ldr.php?AT=pb&SP=MC&rID=107355102&rKey=f738a3c6741624b7

Part 3:  https://datamining.webex.com/datamining/ldr.php?AT=pb&SP=MC&rID=107355112&rKey=0a11055c91cd0439