Automatically assigned DDC number: 00631

Manually assigned DDC number: 00631

Title: Inductive Learning by Selection of Minimal Complexity Representations

Author:

Author:

Subject: Arlindo Manuel,Limede Oliveira Inductive Learning by Selection of Minimal Complexity Representations

Description: Inductive Learning by Selection of Minimal Complexity Representations by Arlindo Manuel Limede de Oliveira Doctor of Philosophy in Engineering in Electrical Engineering and Computer Science University of California at Berkeley Professor Alberto Sangiovanni-Vincentelli, Chair This dissertation addresses the problem of inferring accurate classification rules from examples. A formalization of Occam's razor, the minimum description length principle, is used to transform the problem of performing accurate induction from examples into the problem of selecting the minimal complexity rule that fits well the available data. Four different representation schemes are addressed: two-level threshold gate networks, multilevel Boolean networks, decision graphs and finite state machines. Heuristic algorithms for the inference of classification rules represented using each one of the first three representations are presented and their performance evaluated, both in terms of the size of the solution obt...

Contributor: The Pennsylvania State University CiteSeer Archives

Publisher: unknown

Date: 1994-12-20

Pubyear: 1994

Format: ps

Identifier: http://citeseer.ist.psu.edu/140227.html

Source: http://www-cad.eecs.berkeley.edu/HomePages/aml/publications/report.ps.gz

Language: en

Rights: unrestricted