Learning (Complex) Structural Descriptions from Examples

SUMMARY We present a formalization of an intuitively sound strategy for learning a description from examples : within a partition examples are grouped according to greatest resemblances and examples not in the same subset show a maximum of differences. I. INTRODUCTION WINSTON [4] has demonstrated the importance of the near-miss concept in a context of learning descriptions from examples. His methodology is practical when a few simple scenes are dealt with. We have extended it to include numerous complex examples. The definition of the near miss concept [3 ,4] will be summarized in section 3. A problem arises from the fact that a large number of near-misses can be obtained which do not convey the same type of information. Our experience shows that at least three types of near-misses must be introduced : highly ambiguous, ambiguous and discriminant near-misses, each conveying a different type of information. A second problem concerns the building of a structural description when several examples of several concepts are given, i.e. there are so many possible descriptions that a choice must be made of the most suitable as a recognition device. This leads us to define "promising" partitions of a set of examples. Given a set of examples, the description will be a tree recursively constructed by the rule : divide the set into its most promising partitions. An example of our methodology is given below. This example is actually too simple for the system and must be considered only as an illustration of the definitions we propose.