seahailang的博客

Introduction考虑如下一个随机过程给定分布G0G_0G0​，一个正实数α\alphaα从分布G0G_0G0​中采样X1X_1X1​n>1n>1n>1时：以概率αn−1+α\frac{\alpha}{n-1+\alpha}n−1+αα​ 从G0G_0G0​中采样XnX_nXn​以概率1n−1+α\frac{1}{n-1+\alpha}n−1+.

高斯混合模型假设数据产生过程如下：z∼Mut(π)z \sim Mut(\pi)z∼Mut(π)x∣z∼N(μz,σz2)x|z \sim N(\mu_z,\sigma_z^2)x∣z∼N(μz​,σz2​)记θ={π,μ1,...,μK,σ1,...,σK}\theta = \{\pi,\mu_1,...,\mu_K,\sigma_1,...,\sigma_K\}θ={π,μ1​,......