Mathematical Programming Applications in Pattern Recognition.
Abstract
Problems in pattern recognition are treated by the methods of mathematical programming. In particular the two-class pattern classification model with decision rules based on discriminant functions is considered with emphasis on mathematical programs that determine linear and piecewise linear discriminants. For linearly separable pattern sets of separating hyperplane can be determined by solving a system of linear inequalities. This system serves as the constraint set for a class of mathematical programs that define separating linear discriminants exhibiting maximum tolerance to pattern noise. Specific cases that can be modelled as linear and quadratic programs are discussed and a reliability interpretation of the objective criterion is given.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1978
- Accession Number
- ADA061496
Entities
People
- Robert Hemstreet Leary
Organizations
- Stanford University