第一的，$x_{ij}$不是伯努利随机变量。在文中，$x_{ij}$被描述为实值。此外，尺寸$z$和$w$不正确。

文本表明我们正在绘制一个数据矩阵调用它$X (n\times d)$ 这样 $X_{i,j}=x_{ij}$.

特定条目具有分布 $x_{ij} \sim f(\alpha_{z_i, w_j})$. 在哪里$\alpha$ is some given matrix of parameters. The maximum row index of $\alpha$ is $g$. The maximum column index of $\alpha$ is $m$. So $\alpha$ is $g \times m$.

All that remains is to pin down the indices $z$ and $w$. These are vectors rather than matrices since they have a categorical distribution, for example $z_i$ should be an integer between $1$ and $g$.

Putting this together we need to:

1. Draw $z$, a vector of length $n$, by $z_i \sim \text{Cat}(\pi), i=1,...,n$
2. Draw $w$, a vector of length $d$, by $w_j \sim \text{Cat}(\rho), i=1,...,d$
3. Draw $X$, an $(n\times d)$ matrix, by $x_{ij} \sim f(\alpha_{z_i, w_j})$