Friedman 9704431 A predictive learning system is a computer program that constructs rules for predicting values of some property of a real system ("output") given the values of other properties ("inputs") of that system. Predictive learning systems attempt to construct or "learn" useful prediction rules purely by processing data taken from past successfully solved cases; that is, cases for which the values of both the output and input variables have been determined. The role of the learning algorithm is to automatically extract and organize the information in this data to obtain accurate rules for predicting output values of new cases in which only the values of the input variables are known. This research concentrates on new paradigms for the development of predictive learning methodology quite different from past approaches. The goal is to produce faster more powerful learning algorithms. These new algorithms can then be applied to solve broader classes of more complex problems, such as those that arise in the fields of communications, biotechnology, environmental forecasting, and manufacturing process control. The research will involve making advances in the areas of statistical theory, algorithm design, and high performance computing.
