A maxentropic procedure for reconstructing a distribution from its marginals

We present a maxentropic approach of reconstructing a discrete joint probability distribution pij from the knowledge of its marginals. The version we develop is quite useful when preliminary information is known about the values of some or all the pijʹs.