23.02.2015 Views

Machine Learning - DISCo

Machine Learning - DISCo

Machine Learning - DISCo

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.

10.2.1 General to Specific Beam Search<br />

One effective approach to implementing LEARN-ONE-RULE is to organize the hypothesis<br />

space search in the same general fashion as the ID3 algorithm, but to<br />

follow only the most promising branch in the tree at each step. As illustrated in the<br />

search tree of Figure 10.1, the search begins by considering the most general rule<br />

precondition possible (the empty test that matches every instance), then greedily<br />

adding the attribute test that most improves rule performance measured over<br />

the training examples. Once this test has been added, the process is repeated by<br />

greedily adding a second attribute test, and so on. Like ID3, this process grows the<br />

hypothesis by greedily adding new attribute tests until the hypothesis reaches an<br />

acceptable level of performance. Unlike ID3, this implementation of LEARN-ONE-<br />

RULE follows only a single descendant at each search step-the attribute-value<br />

pair yielding the best performance-rather than growing a subtree that covers all<br />

possible values of the selected attribute.<br />

This approach to implementing LEARN-ONE-RULE performs a general-tospecific<br />

search through the space of possible rules in search of a rule with high<br />

accuracy, though perhaps incomplete coverage of the data. As in decision tree<br />

learning, there are many ways to define a measure to select the "best" descendant.<br />

To follow the lead of ID3 let us for now define the best descendant as the one<br />

whose covered examples have the lowest entropy (recall Equation f3.31).<br />

The general-to-specific search suggested above for the LEARN-ONE-RULE algorithm<br />

is a greedy depth-first search with no backtracking. As with any greedy<br />

IF<br />

THEN PlayTennis=yes<br />

IF Wind=strong t IF Humidity=high<br />

THEN PlayTennis=no IF Hum'ditv=norntal THEN PlayTennis=no<br />

THEN PlayTennis=yes<br />

A/\ - IF Humidity=nowl ...<br />

Wind=strong IF Humidity=normal Outlook=rain<br />

IF Humidify=normal<br />

Wind=weak<br />

THEN PlayTennis=yes IF Humidity=normal<br />

THEN PlayTennis=yes Outlook=sunny THEN PlnyTennis=yes<br />

THEN PlayTennis=yes<br />

FIGURE 10.1<br />

The search for rule preconditions as LEARN-ONE-RULE proceeds from general to specific. At each<br />

step, the preconditions of the best rule are specialized in all possible ways. Rule postconditions are<br />

determined by the examples found to satisfy the preconditions. This figure illustrates a beam search<br />

of width 1.

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

Saved successfully!

Ooh no, something went wrong!