Average Prior Distribution of All Possible Probability Density Distributions

摘要: Bayes’ rule is universally applied in artificial intelligence and especially Bayes reasoning, networks, decision–making, generating rules for probabilistic knowledge bases. However, its application requires about a priori distribution of probability or density that frequently not given. Then, to find at least an approximate solution problem, the uniform used. Do we always have use this distribution? The paper shows it true. prior should only be used if there no real distribution. If however, possess certain qualitative knowledge, e.g. unimodal one, expected value less than 0.5, then can apply being average all possible distributions, instead As result will usually get better approximation problem avoid large errors. explains concept distributions how they determined with special method granulation diminution elementary events probability.