Optimal and Heuristic Synthesis of Hierarchical Classifiers.
Abstract
Multistage schemes such as hierarchical classifiers have been found useful for many multiclass pattern recognition tasks. This dissertation investigates the theoretical properties of a general model of multistage multiclass recognition schemes. The generality of the model allows one to describe a large class of parametric and non-parametric schemes commonly used in terms of the model parameters. Hierarchical classifiers are special types of multistage recognition schemes wherein at each stage certain classes are rejected from consideration as labels of the test sample. Theoretical properties of decision trees whose node decisions are statistically independent are investigated. Even under this independence assumption the optimal tree design task is a complex one.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1976
- Accession Number
- ADA042161
Entities
People
- Ashok Vasant Kulkarni
Organizations
- University of Maryland