抄録
This paper is concerned with a new approach to estimating probability density function when several low-order moments or cumulants are known. The problem is to find an algorithm efficient for estimating probability density functions under the policy to maximize the entropy involved. In deriving this technique low-order moments, which are beforehand determined by any conventional procedures, are put as constraints. Estimation is then reduced to a problem to solve a system of nonlinear equations, whose resolution is performed by the Newton-Raphson method. The main features of this procedure are as follows: Estimation can be made using a rather simple system of equations and is significantly effective in applying to non-Gaussian distributions. Finally, as an example, this method is applied to reactor noise data and results obtained are in good agreement with those by histogram.
