The Principle Of Maximum Entropy And Minimum Entropy Methods In The Measurement Data Processing Applications

Most of Modern Science Researches depend a lot on measurement. There are two important steps in measurement process, which are measurement data obtaining and measurement data processing. Based on the theory of measurement data processing, the applications of Maximum Entropy Method and Minimum Entropy Problem in data processing are researched and a new application method is explored in this paper. The new method is called MinMax Information Measure, which is used to deal with the actual measurement data.In the analysis of measurement data with adulterated subjective factor, it is difficult to accurately reflect objective measurement result. To solve this problem, the maximum entropy method (MEM) is used to determine the probability distribution of the given measurement data, and the probability distribution is estimated and evaluated based on the different moments constraints through the maximum entropy method. The MinMax Information Measure can measured the moment constraints which contain the most information, and can analysis the relation between the sample data and moment constraints. So we can identify the sample moment constrains which is closest to the real distribution.In the actual process of measurement, a automation test system is built and frequency data is obtained to validate the methods presented in this paper. At present, the frequency data of measurements is a physical which copy of the most accurate, maintain the most stable and measure the most accurate. The simulation and calculation results prove that the probability distribution determined by the MEM is the reasonable distribution with the least assumption, and the probability distribution obtained would be closer to the real distribution as the moment constrains are increased. The Minimum Entropy Problem was solved through the method of Hopfield neural network and simulated annealing. At last, The moment constraints which contain most information can be computed and identified the differences between additional moment constraints through the MinMax Information Measure, so the character moment constraints of small sample data which contain the most information can be identified.