The Maximum Entropy Model Of Fuzzy T-S

For all kinds of models in the field of NLP, the maximum entropy model is form the simplest and implement one of the most complex models. Maximum entropy model is the basic idea is: given the training sample, choose a model consistent with training samples, the maximum entropy model is consistent with these observations should be selected for the probability distribution, and in the case of besides, model with uniform probability distribution. As the known conditions for maximum entropy probability model, it is the biggest difficulty comes from feature selection, parameter estimation and calculation of the normalization factor. Usually the maximum entropy model of the solution is to use a general iterative algorithm called GIS or IIS. But due to the slow convergence speed, easy crossing the line, and still a huge amount of calculation, thus the maximum entropy model has not been widely used. Along with the development of fuzzy mathematics, and the successful application in many fields,so this paper will establish t-s fuzzy model of the maximum entropy model, Logistic model after using transform as the maximum entropy of fuzzy differential equation,the control variables instead of normalized factor, maximum entropy global fuzzy model is established, using MATLAB simulation results. And compared with the traditional GIS algorithm to do the comparison of the accuracy and response speed, in terms of accuracy, GIS method and maximum entropy fuzzy control are basically the same; In terms of response speed, when large amounts of data, t-s fuzzy control than the GIS algorithm obtained good effect. As a result, the maximum entropy model of t-s fuzzy control in the era of big data will have a better development, in the field of natural language and machine learning will take advantage of wide application. This paper mainly study the calculation of the maximum entropy model, realize the big data under the background of rapid calculation. Opens a new way to solve the problem of classical maximum entropy, also found that can be controlled by control characteristic variables of the maximum entropy as a result, the reverse input/output control of the data has certain enlightening significance.