1: \begin{abstract}

2: Importance weighted variational inference \citep{burda2015importance} uses multiple i.i.d. samples to have a tighter variational lower bound. 

3: We believe a joint proposal has the potential of reducing the number of redundant samples,

4: and introduce a hierarchical structure to induce correlation.

5: The hope is that the proposals would coordinate to make up for the error made by one another to reduce the variance 

6: of the importance estimator.

7: Theoretically, we analyze the condition under which convergence of the estimator variance can be connected to convergence of the lower bound.

8: Empirically, we confirm that maximization of the lower bound does implicitly minimize variance.

9: Further analysis shows that this is a result of negative correlation induced by the proposed hierarchical meta sampling scheme, and performance of inference also improves when the number of samples increases.

10: \end{abstract}

11: