DID YOU KNOW? DTIC has over 3.5 million final reports on DoD funded research, development, test, and evaluation activities available to our registered users. Click HERE
to register or log in.
AN ABSTRACT MODEL FOR THE PATTERN RECOGNITION PROCESS,
MICHIGAN UNIV ANN ARBOR COOLEY ELECTRONICS LAB
The goal of this work, unlike that of other work in the field of pattern recognition, is the construction of a general recognition model which can apply to any group of patterns and to any amount of initial knowledge about them. In the preliminary model which has been developed, the initial knowledge is presented to the recognizer as a list of conditional probabilities then, given certain characteristics of the sequence of items which has been seen, the most probable correct answer can be selected. Where these conditional probabilities are not known exactly, they are estimated by a primitive learning process. Decisions are based on a posteriori probabilities and on other criteria studied in decision theory. The model is described in precise detail, and some possible goals of further work are mentioned. Author
AF49 638 884