Similarity and entropy
Richard V. Smith · 1964
There is much interest today in inductive processes. This typically involves the necessity to classify objects through some sort of pattern recognition without benefit of any previously given classification rules. Either a learning approach or a direct approach may be taken. When a direct approach is used, the question of similarity among the objects arises and a measure for similarity is required. In this paper we will suggest such a measure to give the similarity of objects in a given set relative to a given set of properties. Measures of this kind have previously been proposed but none adequately consider the effect of dependency among the properties. The measure we will develop has its roots in the measurement of entropy. Therefore, we will begin with a brief summary of this topic from information theory.