Review of the Hastie and Tibshirani course: Statistical Learning and Data Mining

I just got back from a two day short course by Trevor Hastie and Robert Tibshirani based on their Stanford graduate stats course and book: Elements of Statistical Learning. The first edition of this book has been a near constant companion in my career, and the new second edition has some notable additions, including new sections on Random Forests, and Elastic Nets, which combines some of the elements of ridge and lasso regression. H&T generously give away their book in PDF form at the book and course page (http://www-stat.stanford.edu/~tibs/ElemStatLearn).
The course was held at the Georgetown University Hotel and Conference Center. Registration for the course was

$1300 for industry attendees. Less for students and faculty. The fee gets you two full days of instruction by Hastie and Tibshirani themselves, a copy of their second edition text, and full course notes. A nice lunch was provided each day, with several coffee breaks with finger foods and plenty of beverages. So while steep, the price is well worth it to me.
The course is structured into roughly equal length parts that they alternate presenting.
Topics were

• Introduction
• Cross validation and bootstrap
• Linear and logistic regression
• Lasso and other penalization methods
• Comparison of LARS, forward stagewise regression, and lasso
• CART and PRIM
• Dimension reduction (SVD, PCA, ICA)
• Trees - Bagging, Boosting, Random Forests
• Sparse graphical models
• Feature selection, False Discovery Rate
• Kernel Methods

The treatment of the methods was solidly aimed at practitioners. Almost no time was spent on theories, proofs, or lemmas. It was all application. Each method was tackled roughly in the order of motivation, application, best use, and examples. There were many examples throughout the course, mostly replicated from the book. At times it felt like they were rushing through some topics where i would have liked to go deeper into the particulars of the technique, but it seemed the majority of the audience liked the pace.

Another nice thing about Hastie and Tibshirani's work is that they have many packages for R that they've contributed to CRAN for all our use. It's no secret I'm an R fanatic, so this is a big plus and just shows how they are such stand-up guys - they could have easily profited from their work by collaborating with a software vendor, but chose to give their work to the community that supports them.... goose bumps....

I highly recommend this course, and if you can make it, you'll be well rewarded by meeting and learning from two great contributors to the statistics and machine learning field.