Sciweavers

ICML
1994
IEEE

Greedy Attribute Selection

14 years 3 months ago
Greedy Attribute Selection
Many real-world domains bless us with a wealth of attributes to use for learning. This blessing is often a curse: most inductive methods generalize worse given too many attributes than if given a good subset of those attributes. We examine this problemfortwolearningtasks takenfroma calendar scheduling domain. We show that ID3/C4.5 generalizes poorlyon these tasks if allowed to use all available attributes. We examine five greedy hillclimbing procedures that search for attribute sets that generalize well with ID3/C4.5. Experiments suggest hillclimbing in attribute space can yield substantial improvements in generalization performance. We present a caching scheme that makes attribute hillclimbingmore practical computationally. We also compare the results of hillclimbing in attribute space with FOCUS and RELIEF on the two tasks.
Rich Caruana, Dayne Freitag
Added 27 Aug 2010
Updated 27 Aug 2010
Type Conference
Year 1994
Where ICML
Authors Rich Caruana, Dayne Freitag
Comments (0)