Entropy-metric estimation of the small data models with stochastic parameters

Area of Science:

• Statistics
• Mathematical Modeling
• Information Theory

Background:

• Formalizing dependencies in datasets, especially small data, is crucial.
• Hypotheses about data properties are key to accurate modeling.
• Existing methods struggle with limited and noisy small datasets.

Purpose of the Study:

• To formalize optimal estimation of probability density functions for parameters in dynamic and static small data models.
• To develop methods for linear and nonlinear models incorporating specific object property hypotheses.
• To address the challenge of parameter estimation with limited, censored, and noisy measurements.

Main Methods:

• Probability theory and mathematical statistics.
• Information theory and evaluation theory.
• Stochastic mathematical programming and information entropy maximization.

Main Results:

• Developed a mathematical framework based on maximizing information entropy for small data.
• Formalized linear and nonlinear dynamic and static small data models with stochastic parameters.
• Successfully determined optimal estimates for probability density functions of model parameters.

Conclusions:

• The formalized procedure provides optimal parameter estimates for small data models.
• The approach effectively handles censored and noisy measurements by maximizing information entropy.
• Optimization problems are reducible to canonical forms for stochastic linear programming.