18.11.2012 Views

anytime algorithms for learning anytime classifiers saher ... - Technion

anytime algorithms for learning anytime classifiers saher ... - Technion

anytime algorithms for learning anytime classifiers saher ... - Technion

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>Technion</strong> - Computer Science Department - Ph.D. Thesis PHD-2008-12 - 2008<br />

List of Tables<br />

3.1 Possible training set <strong>for</strong> <strong>learning</strong> the 2-XOR concept . . . . . . . . 21<br />

3.2 Characteristics of the datasets used . . . . . . . . . . . . . . . . . 46<br />

3.3 The size of the induced trees on various datasets . . . . . . . . . . 49<br />

3.4 The differences in the size of the induced trees on various datasets 50<br />

3.5 The generalization accuracy of the induced trees on various datasets 51<br />

3.6 The differences in generalization accuracy of the induced trees on<br />

various datasets . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52<br />

4.1 Characteristics of the datasets used to evaluate ACT . . . . . . . 82<br />

4.2 Average cost of classification <strong>for</strong> different mc values . . . . . . . . 85<br />

4.3 DTMC vs. ACT and ICET vs. ACT using statistical tests . . . . 86<br />

4.4 Average cost of classification <strong>for</strong> mc = 100 . . . . . . . . . . . . . 87<br />

4.5 Average cost of classification <strong>for</strong> mc = 500 . . . . . . . . . . . . . 88<br />

4.6 Average cost of classification <strong>for</strong> mc = 1000 . . . . . . . . . . . . 89<br />

4.7 Average cost of classification <strong>for</strong> mc = 5000 . . . . . . . . . . . . 90<br />

4.8 Average cost of classification <strong>for</strong> mc = 10000 . . . . . . . . . . . . 91<br />

4.9 Average cost when test costs are assigned randomly . . . . . . . . 93<br />

4.10 Comparison of various <strong>algorithms</strong> when error costs are nonuni<strong>for</strong>m 94<br />

5.1 Characteristics of the datasets used to evaluate TATA . . . . . . . 112<br />

5.2 Comparing the misclassification cost <strong>for</strong> different contracts . . . . 113<br />

B.1 Testing Occam’s empirical principle <strong>for</strong> consistent trees . . . . . . 149<br />

B.2 Testing Occam’s empirical principle when the trees are pruned . . 151

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!