homework1b solutions

867 machine learning fall 2006 mit opencourseware

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: ng XiAlpha-estimates...done Runtime for XiAlpha-estimates in cpu-seconds: 0.02 Cite as: Tommi Jaakkola, course materials for 6.867 Machine Learning, Fall 2006. MIT OpenCourseWare (http://ocw.mit.edu/), Massachusetts Institute of Technology. Downloaded on [DD Month YYYY]. XiAlpha-estimate of the error: error<=0.65% (rho=1.00,depth=0) XiAlpha-estimate of the recall: recall=>99.48% (rho=1.00,depth=0) XiAlpha-estimate of the precision: precision=>99.30% (rho=1.00,depth=0) Number of kernel evaluations: 164632 Writing model file...done Since this is a hard-margin SVM, if a solution exists then the training error will be 0 (and you get away without submitting an image). With the new model file, we can evaluate our model on the test set: $ svm_classify test-01-images.svm svm_model Reading model...OK. (84 support vectors read) Classifying test examples..<snip>..done Runtime (without IO) in cpu-seconds: 0.00 Accuracy on test set: 99.91% (2113 correct, 2 incorrect, 2115 total) Precision/recall on test set: 99.91%/99.91% Now, for interest’s sake, we’d like to extract the two test images that were misclassified. This is how it’s done using unix power tools: $ cat test-01-images.svm | awk ’{ print $1 }’ | paste - svm_predictions \ | nl | awk ’{ if ($2 * $3 < 0) { print $1 } }’ 1665 2032 It can also be done somewhat less easily by using a spreadsheet utility. Here are the images: test-001665.png test-002032.png We also accepted training with the default value of C : $ svm_learn train-01-images.svm Scanning examples...done Reading examples into memory...<snip>..OK. (126...
View Full Document

This document was uploaded on 03/20/2014 for the course EECS 6.867 at MIT.

Ask a homework question - tutors are online