Semi-Amortized Variational Autoencoders
This incremental method addresses posterior collapse in VAEs for text generation, benefiting researchers in deep generative modeling.
The paper tackles the problem of suboptimal variational parameters in amortized variational inference for VAEs by proposing a semi-amortized approach that initializes parameters with AVI and refines them with differentiable SVI, resulting in improved performance over baselines on text and image datasets.
Amortized variational inference (AVI) replaces instance-specific local inference with a global inference network. While AVI has enabled efficient training of deep generative models such as variational autoencoders (VAE), recent empirical work suggests that inference networks can produce suboptimal variational parameters. We propose a hybrid approach, to use AVI to initialize the variational parameters and run stochastic variational inference (SVI) to refine them. Crucially, the local SVI procedure is itself differentiable, so the inference network and generative model can be trained end-to-end with gradient-based optimization. This semi-amortized approach enables the use of rich generative models without experiencing the posterior-collapse phenomenon common in training VAEs for problems like text generation. Experiments show this approach outperforms strong autoregressive and variational baselines on standard text and image datasets.