| Entropy as a basic concept in information theory has been widely concerned and studied. The principle of maximum entropy based on entropy has been developing in the research and application of many scholars and has gradually formed its own theoretical system, and has been widely used. For example, when the Bayesian method is used to evaluate the measurement uncertainty, the principle of maximum entropy can estimate the prior distribution. In this paper, Shannon entropy is taken as the main object, and the maximum entropy principle under different constraints involved in the current research is summarized, and sorted out into a relatively unified model. This paper discusses some methods to improve the traditional maximum entropy principle by using transformation function method and density kernel estimation method, and introduces their improvement ideas, theoretical models and application characteristics in detail. Finally, combined with practice, the improved methods of maximum entropy principle are summarized from the aspects of constraint selection, evaluation index and optimization algorithm, which will promote the further research and application of maximum entropy principle. |
