| Popis: |
The probability distribution of data are parameter-based and typically formatted as script numbers. However, these values areusually estimated experimentally, leading to selecting only one value is error-prone. Moreover, the real data often follows amixed form of probability distributions, in which sub-datasets are often incomplete data. It is possible to use parameters forensuring flexibility, especially in classification problems. Therefore, in this paper, we propose a new method for describingparameters estimated through Bayesian statistics; the new similarity between two probability distributions with the fuzzyparameters, through the fuzzy extended Kullback – Leibler divergence. The paper also applied the proposed methods in leaf classification. |
