Variational Bayesian Inference for Logistic mixture model
The traditional variational Bayesian inference method fails to perform effective parameter estimation when inferring non-conjugate mixed models.Therefore,a new variational Bayesian algorithm is proposed.This algorithm solves the problem of not forming conjugate distribution by approximating the Logistic function,and then realizes the parameter estimation of this kind of mixed model.Firstly,the component is fixed and the prior setting is given.Then,the complete derivation of the variational Bayesian inference is given,and on this basis,the Variational Bayesian Inference algorithm is proposed to solve the problem of parameter estimation of the mixed model.Finally,an empirical analysis is carried out on the dataset.The corresponding and the results show that the algorithm can achieve fast convergence for its accurate estimation.
NETL
NSTL
万方数据
