On an approach to tuning the Metropolis–Hastings algorithm for the task of separating a mixture of Gaussian components

The article considers the problem of separating a mixture of Gaussian components, which consists in determining, from available observations, the parameters of the mixture components. An approach to solving this problem is proposed, based on Bayesian estimation using the most informative prior distributions (Maximal Data Information Prior – MDIP). The novelty of the described approach lies in the use of sample estimates to calculate the prior distribution and determine the settings of the Metropolis-Hastings algorithm for sampling with adaptive stepwise adjustment of the proposed distribution parameters.