[Proceedings] Third International Conference on Tools for Artificial Intelligence - TAI 91
DOI: 10.1109/tai.1991.167073
|View full text |Cite
|
Sign up to set email alerts
|

Data-driven constructive induction in AQ17-PRE: A method and experiments

Abstract: This paper presents a method for constructive induction, in which new attributes are constructed as various functions of original attributes. Such a method is called data-driven constructive induction, because new attributes are derived from an analysis of the data (examples) rather than the generated rules. Attribute construction and rule generation is repeated until a termination condition, such as the satisfaction of a rule quality measure, is met. The first step of this method, the generation of new attrib… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
7
0

Publication Types

Select...
6
1

Relationship

0
7

Authors

Journals

citations
Cited by 17 publications
(7 citation statements)
references
References 4 publications
0
7
0
Order By: Relevance
“…If an adequate model does not lie within that space, then it is not possible for the machine learning system to find one. While constructive induction [17][18][19][20][21][22][23][24][25][26][27][28][29][30][31] may construct new complex terms from given primitive terms, it is unable to construct new terms that can only be described using terms that lie outside the given vocabulary. By contrast, human experts are accomplished at extending and refining metamodels as circumstances demand and they have ready access to multiple sources of insight that may help guide such a process.…”
Section: Relative Capabilities Of Machine Learning and Knowledge Acqumentioning
confidence: 99%
“…If an adequate model does not lie within that space, then it is not possible for the machine learning system to find one. While constructive induction [17][18][19][20][21][22][23][24][25][26][27][28][29][30][31] may construct new complex terms from given primitive terms, it is unable to construct new terms that can only be described using terms that lie outside the given vocabulary. By contrast, human experts are accomplished at extending and refining metamodels as circumstances demand and they have ready access to multiple sources of insight that may help guide such a process.…”
Section: Relative Capabilities Of Machine Learning and Knowledge Acqumentioning
confidence: 99%
“…A di erent formalism, decision lists, i n w h i c h the rule ordering is signi cant, will be examined later. Such rules can beconstructed by generalising from positive and negative examples of a concept using the Candidate Elimination algorithm 71] (AQ 66,11,12] The AQ algorithms start with a random single instance (seed) and form the most general hypothesis which includes no negative examples by successively specialising an initially empty complex, using tests that incorporate the seed. Geometrically, they obtain the biggest hyperrectangle enclosing no negative points.…”
Section: Rule-based Systemsmentioning
confidence: 99%
“…It is somehow connected with postulates of a constructive induction to the change of the input data (cf. [2]). …”
Section: Multistrategic Approach To Determining the Relevant Attributesmentioning
confidence: 99%