Home

About

Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network
FAQHELPSITE MAPCONTACT US


  Advanced Search  

 
IEEE TRANSACTIONS oN SYSTEMS, MAN' AND CYBERNETICS-PAM B: CYBERNETTCs, vol. 28, No. 2, ApRIL l99E Nonparametric Discriminant Analysis via
 

Summary: IEEE TRANSACTIONS oN SYSTEMS, MAN' AND CYBERNETICS-PAM B: CYBERNETTCs, vol. 28, No. 2, ApRIL l99E
Nonparametric Discriminant Analysis via
Recursive Optirnization of Patrick-Fisher Distance
Mayer E. Aladjem
Abstracr- A method for the linear discrimination of two classes ispresented. It searches for the discriminant direction which maximizes
the Patrick-Fisher (PF) distance between the projected class-conditional
densities. It is-a nonparametric method, in tlie J"rrr. that the densities
are estimated from the data. Since the PF distance is a highly nonlinear
functionr w propose a recursive optimization procedur. roi searching
the directions corresponding to several large local maxima of the pF
distance' rts novelty lies in the transformation of the data along a founddirection into data with deflated maxima of the pF distance and iterationto obtain the next direction. A simulation study
"na
a medical dataanalysis indicate the potentiat of the method to nna the ,.qu.rr.. ofdirections with significant crass separations.
I. IvrnooucrroN
We discuss discriminant analysis of two classes which is carried
out by the linear mapping r - r?x, x Rn, r e Rr, n) 2, with
x an arbitrary n-dimensional observation, and r a direction vector
(having unit length). The vector r maximizes the patrick-Fisher
(PF) distance [6] which measures the overlap of the class-conditional
densities along r. Unfortunately, the PF distance is not a unimodal

  

Source: Aladjem, Mayer - Department of Electrical and Computer Engineering, Ben-Gurion University

 

Collections: Computer Technologies and Information Sciences