No abstract

Pattern recognition is a newly developing branch of artificial intelligence that shows a great deal of promise in providing a generalized approach to solutions of a large class of data analysis problems in experimental chemistry. A general statement of the problem is: can an obscure property of a collection of objects (elements, compounds, mixtures, etc.) be detected and/or predicted using indirect measurements made on the objects? One particular method within the realm of pattern recognition, the learning machine, has been successfully applied to spectroscopic data for direct detection of molecular structural units. This paper introduces pattern recognition in a much broader scope. Using a synthetic data base and a data base of chemical interest, the major approaches within pattern recognition are examined. One method representing each approach is applied to the two fundamentally different data sets, first to compare the results, but also to illustrate the far-reaching problem solving capability.

No abstract

No abstract

No abstract