AN APPROACH TO DATA-DRIVEN LEARNING

In the present paper a data-driven approach to learning is described. The approach is discussed in the framework of the Net-Clause Language (NCL), which is also outlined. NCL is aimed at building network models and describes distributed computational schemes. It also exhibits sound semantics as a data-driven deductive system. The proposed learning scheme falls in the class of methods for learning from examples and the learning strategy used is instance-to-class generalization. Two basic examples are discussed giving the underlying ideas of using NCL for inductive concept learning and learning semantic networks.