| 摘要 |
The present invention discloses a mixed model determination method and apparatus and belongs to the field of statistics. The method comprises: acquiring a logarithm likelihood, a regular term and logarithms of variation distribution of hidden variables, which are determined according to sample data, the hidden variables and component parameters, and according to the logarithm likelihood, the regular term and the logarithms of the variation distribution of the hidden variables, determining a target function; and according to sub-sample data extracted from the sample data, determining the variation distribution of the hidden variables and the component parameters, which enable the target function to be converged, and according to the variation distribution of the hidden variables and the model parameters, which enable the target function to be converged, determining a mixed model. According to the present invention, the target function is determined by the logarithm likelihood, the regular term and the logarithms of the variation distribution of the hidden variables, which are determined according to the sample data, the hidden variables and the component parameters, and the mixed model is determined according to the variation distribution of the hidden variables and the component parameters, which enable the target function to be converged; due to introduction of the regular item, complexity of the mixed model is automatically controlled; and the mixed model determination method and apparatus can be suitable for processing a super-large scale of data. |
