A model geometry for the HIV-1 and -2 viruses is proposed based on electron microscopic observations. Standard geometric shapes are employed to describe the surface membrane, stalk (glycoprotein 41), and knob (glycoprotein 120) of the virus. Based on approximate dimensions for certain features of the virus estimates are provided for viral surface area and volume in both budding and mature stages, stalk-knob assembly surface area and volume, surface membrane area, surface membrane radius, and the ratio for surface area in the budding and mature virus. Such estimates may be useful in the quest for compounds that inhibit virus replication.

In the past decade, there has been some interest shown in the use of linear programs to solve discriminant analysis problems. In this paper we present and clarify an alternative to the statistical approach for discriminating between two (or more) groups of vector-valued data in which an order relation is given.

By the nature of this problem, a bicriteria linear program can be used instead of a statistical approach and hence no statistical assumptions need to be imposed on the problem.

The well known "SOR" method is obtained from a one-part splitting of the system matrix $A$, using one parameter $\omega$.

M. Sisler introduced a new method by using one parameter for the lower triangular matrix $L$. Later he combined the above two methods to get a two parametric method [7], [8], and [9].

D. Young considered yet another two parametric method. The two parameters weight the diagonal of a positive-definite and consistently ordered 2-cyclic matrix [6]. Removing Young's hypothesis that both parameters are in the interval $(0,1]$, we generalized his theorem.