Takeo Yukimachi

On measure theoretical structure of multivariate information among random variables

This paper is concerned with a quantity called “multivariate information” among random variables. By defining it as a specialization of the information proposed by Kullback and Leibler, we will discuss fundamental properties of it, which are direct applications of the results of mathematical information theory. We will derive some formulas about multivariate information closely related to the absolute continuity of the joint probability distribution with respect to some other distribution.

A method for system reliability analysis with change-over operation

Abstract When fault tree analysis (FTA) is used in analysing the system reliability including human functions, the correspondence between an abnormal event and the human reaction against it may become ambiguous. In order to supplement this defect, a case study has been conducted in this research to develop a new technique termed ‘Corrective Operation Diagram Analysis’, and an attempt has been made to establish the correspondence between the reliability analysis of equipment/hardware using FTA and human reliability analysis.

Multivariate information rate

In this paper, a new quantity called “multivariate information rate” is introduced for discrete-parameter random process. This quantity is an average multivariate information included in a given process, and we discuss some fundamental properties of it mainly for the case of stationary Gaussian random sequences. On the basis of such properties, we evaluate the multivariate information rates of random sequences that appear in a data processing system. Practically, the notion of this quantity seems to be useful in the situation of discrete smoothing.