1: \begin{abstract}

2: We introduce a new algorithm for approximate inference that combines reparametrization, 

3: Markov chain Monte Carlo and variational methods. We construct a very flexible implicit variational distribution 

4: synthesized by an arbitrary  Markov chain Monte Carlo  operation and a deterministic  

5: transformation that can be optimized using the reparametrization

6: trick. Unlike current methods for implicit 

7: variational inference, our method avoids the computation of log density ratios and therefore it is 

8: easily applicable to arbitrary continuous and differentiable models. We demonstrate the proposed 

9: algorithm for fitting banana-shaped distributions and for training variational autoencoders.   

10: %We will also discuss a second class

11: %of MCMC and variational inference hybrids based on the idea "learn how to start to speed up convergence" 

12: %that is suitable for discrete models and it makes use of the log derivative trick.       

13: \end{abstract}

14: