Sequential pattern recognition

Consider an n-dimensional space which has been partitioned into 't' unique subspaces, called categories or populations. Associated with each category is a set of n-tuples, to be referred to as pattern vectors. Since each pattern vector belongs to one and only one category, each vector may be considered to be a data vector having 'n' dimensions and belonging to a specific category. A program was written which realizes an algorithm that performs an adaptive process with data of known classification. This procedure will establish the necessary criterion for a classification scheme for other data from the same space whose category is unknown. A second program was written which performs the classification process according to the criterion established by the adaptive technique. It should be noted that a priori knowledge about the probability distributions of the data sets need not be known. Several test problems were run on the IBM 360 digital computer using geophysical data. The results of these runs were highly successful in correctly classifying the data that was used.