The two different definitions of the "amount of information", due respectively to N. Wiener and to C. E. Shannon, apparently contradict each other. Although recently L. Brillouin clarified the interrelation between them by a very ingenious argument, there are still several ambiguous points which require further remarks. We intend to make the notion of the amount of information much clearer by extending his arguments and making more profound considerations of these points. In§1 two definitions are compared and a unified interpretation is given. It becomes clear that they are in fact equivalent, merely looking the same thing from different points of view. Both are equal to the decrease of entropy during the process of extracting the message:I=-ΔS. In§2 the Brillouin's argument is somewhat extended, introducing the notion of the noise of the first kind. Ordinary noise, which is called here that of the second kind, is introduced in §3. With due definitions of "message entropy" and "noise entropy", the interpretation made in §1 is found to be not only still valid but able to claim the most generality. Misinterpretations about Shannon's theory can be avoided by introducing these notions of different kinds of entropy. In§4 the relationship between our entropies and those which appear in Brillouin's arguments is discussed, with some miscellaneous remarks.
抄録全体を表示