Some distinct merits has pattern classification using $VL_1$ formula compared with other traditional (especially mathematical) approach.
1. It can deal with problems containing nonlinear or norminal variables.
2. Simplified $VL_1$ classification rule through reduction variables requires less computing time than that in case of mathematical approach to classify input data.
3. Understanding the process of classification is very easy to comprehend because it resambles human tendancy to recognize objects or classes of objects.
Thus it is considered as a powerful tool to solve the problems which have large number of variables or classes and norminal or non-linear variables. The concept of VVL was first introduced in 1972 by R.S. Michalsky[1].
$VL_1$ is a kind of VVL having only one output, which was devised especially for pattern classification. Michalski also developed Algorithm G, covering algorithm for classes experienced in $VL_1$ formula. This thesis is the FORTRAN implementation of $VL_1$ formula synthesis based on the algorithm. The first, and only computer implementation of Algorithm G in $PL_1$ has been made by Larson and Michalski in bit string mode, which works for the problems with less than 11 cardinality of domain [5].
On the other hand, employing the boundary value manipulation in this program makes it possible to deal with problems containing even extremely large size of domain or improperly selected domain without any difficulties.
Thus, it can be applied to various pattern recognition problems, such as speech recognition, feature selection, scene matching and image understanding, etc.