A Basic Model for Learning Systems

Abstract

A learning system (LS/O) is designed and implemented on a computer. The LS/O exchanges information with its environment through three kinds of message strings. Given a question string, the LS/O produces the response string, then the answer string is shown to the LS/O as the correct response string. The LS/O iterates such interactions infinitely. The most important feature of the LS/O is that it tries to produce the correct response to the unexperienced questions. To do this, the LS/O organizes the previously acquired information and generates the hypothesis or the knowledge structure. When the response fails to meet the answer, the LS/O renews the hypothesis or reorganizes the knowledge structure to explain the answer. In the LS/O, the network-like structure called label net plays an important role to represent the knowledge structure.