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 PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 19, NO. 2, FEBRUARY 1997 187 1 lrurnoDucnoN
 

Summary: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 19, NO. 2, FEBRUARY 1997 187
1 lrurnoDucnoN
To obtain a visual representation of high dimensional data, we
*ttst reduce the dim6nsiona bty, and for data visualization two-t.
dimensional representations (scatter plots) are most useful [5],
t13l' In this PaPer we discuss discriminant analysis for tuso clqsses
I12l' Scatter plots intended for discriminant analysis are called
discriminant plots. Our consideration is limited to^ plots obtained by
the lineai mapping y =frr, r2Jrx, x e R', y . *; n > 2 with x an
arbitrary n-dimensional observafion, und tt, rz direction vectors(each having unit length). vectors rr, 12 whicn opti* ize adiscrimi-
nant criterion are called discriminant aectors.
we discuss two crasses of disaiminant criteria, namery, the
extended Fisher criterion previousry proposed by us [1], tzl and
the nonparametric criterion propos"d by Fukunaga tgl. Our goal is
to assess the discrimination q.rutiuu, or the vecior, ,, and 12 ob_
tained by successive optimization of these criteria. In the past, two
methods .for successive optimization were appried. The first
let!9a ry:t-glthogonal constraints on the discriminant vectors. Itis called oRTH-in this Paper. The second method does not so con-
strain the.discriminant vectors. It is called FREE. our experience
shows that method FREE does not always create new discriminant

  

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

 

Collections: Computer Technologies and Information Sciences