| 摘要 |
The invention discloses a method and device for determining a relation model, and belongs to the technical field of statistics. The method includes the steps that an objective function is determined according to at least two sets of sample data, at least three sets of hidden variables, logarithm likelihood and regular terms determined by model parameters and variation distributed logarithms of the hidden variables; the relation model is determined according to variation distribution of the hidden variables converged by the objective function and the model parameters. The objective function is determined according to the at least two sets of sample data, the at least three sets of hidden variables, the logarithm likelihood and the regular terms determined by the model parameters and the variation distributed logarithms of the hidden variables; the relation model is determined according to variation distribution of the hidden variables converged by the objective function and the model parameters, the relation model is determined according to multiple relations between the multiple sets of sample data, the determination efficiency and accuracy of the relation model are improved, the regular terms are introduced, and the complexity of the relation model is automatically controlled. |
