A Note on Computing the Asymptotic Form of a Limited Sequence of Decision Trees.
Abstract
It is often desirable to compute the asymptotic form of a limited sequence of decision trees (DT's) after the set of decision variables has reached a stable, constant membership. Such is the case when the DT's characterize the evolving behavior of a learning strategy in the Quasi-Optimizer system under development. An algorithm is described which builds the asymptotic decision tree in breadth-first order, node-by-node, starting at the root. In the course of computing the asymptotic value of a node (that is, its out-degree and the limiting points of its subranges), three situations may arise: (1) the computation is possible; (2) the computation seems to be possible but more DT's have to be acquired of the evolving strategy to reach the desired level of statistical significance; (3) the computation is not possible. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 1981
- Accession Number
- ADA110254
Entities
People
- Bede Mccall
- Neal Mazur
- Nicholas V. Findler
Organizations
- University at Buffalo