Homework 8

Homework 8#

Deadline

Due Sunday April 6th at midnight on Crowdmark.

This homework covers 3 classes:

  • Mon 3/31, we covered 9.1 - Maximal Margin Classifier

  • Weds 4/2, we covered 9.2 Support Vector Classifiers

  • Fri 4/4, we covered 9.3, 9.4 Support Vector Machines

Problems listed below are from the textbook.

  • 9.7.3 Parts (a) through (f).

    • I’ve created a desmos toy to help you: https://www.desmos.com/calculator/hvfkvby64z

    • Feel free to include marked up screenshots of that tool in your explanation.

    • Please note that it does not provide a unit length vector, so be sure to modify your answer to part (b) accordingly.

  • 9.7.8 Uses the OJ.csv data set included in the DataSets folder. Do all parts, but notes are below.

    • (b) Use C=1 (the default) instead of the book’s C = 0.01 information.

    • (d) Use GridSearchCV like we used in class. You can use C = [0.01, 0.1, 1, 10] in the range. However, we have noticed that particularly for the linear kernel, GridSearchCV can be incredibly slow. If you are having issues getting the results to finish, feel free to use a subset of C values from above, or set cv=5 to use fewer splits in the \(k\)-fold CV.

Important

Standard instructions for submissions and deadlines can be found on the Homework Info Page.