Abstract
Linear decision functions approximating the optimal decision functions in the least-mean-square-error sense are determined. By choosing a particular form of optimal decision functions, it is possible to evaluate the linear decision functions explicitly. It is found that this approach embodies some existing methods for evaluating linear discriminant functions, and is applicable to more general performance criteria. An iterative algorithm which improves the performance of the linear classifier based on linear decision functions substantially is proposed.
Original language | English (US) |
---|---|
Title of host publication | Unknown Host Publication Title |
Pages | 448-452 |
Number of pages | 5 |
State | Published - 1968 |
Externally published | Yes |
Event | Princeton Univ-2nd Annual Princeton Conference on Information Sciences & Systems-Proc - Duration: Mar 25 1968 → Mar 26 1968 |
Other
Other | Princeton Univ-2nd Annual Princeton Conference on Information Sciences & Systems-Proc |
---|---|
Period | 3/25/68 → 3/26/68 |
ASJC Scopus subject areas
- General Engineering