Name
Abstract | ||
|---|---|---|
We prove the following results. Any Boolean function of O(log n) relevant variables can be exactly learned with a set of non-adaptive membership queries alone and a minimum sized decision tree representation of the function constructed, in polynomial time. In contrast, such a function cannot be
exactly learned with equivalence queries alone using general decision trees and other representation classes as hypotheses.
Our results imply others which may be of independent interest. We show that truth-table minimization of decision trees can
be done in polynomial time, complementing the well-known result of Masek that truth-table minimization of DNF formulas is
NP-hard. The proofs of our negative results show that general decision trees and related representations are not learnable
in polynomial time using equivalence queries alone, confirming a folklore theorem.
|
Year | DOI | Venue |
|---|---|---|
1999 | 10.1016/S0020-0190(99)00063-0 | Information Processing Letters |
Keywords | DocType | Volume |
boolean function,dnf formula,general decision tree,folklore theorem,exact learning,representation class,irrelevant variables abound,irrelevant variable,decision tree representation,polynomial time,decision tree,truth-table minimization,related representation | Conference | 70 |
Issue | ISSN | ISBN |
5 | 0302-9743 | 3-540-65701-0 |
Citations | PageRank | References |
8 | 0.56 | 19 |
Authors | ||
3 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| David Guijarro | 1 | 8 | 0.56 |
| Víctor Lavín | 2 | 25 | 3.54 |
| Vijay V. Raghavan | 3 | 2544 | 506.92 |