A Bayesian pseudocoreset is a compact synthetic dataset summarizing essential information of a large-scale dataset and thus can be used as a proxy dataset for scalable Bayesian inference. Typically, a Bayesian pseudocoreset is constructed by minimizing a divergence measure between the posterior conditioning on the pseudocoreset and the posterior conditioning on the full dataset. However, evaluating the divergence can be challenging, particularly for the models like deep neural networks having high-dimensional parameters.

A Bayesian pseudocoreset is a compact synthetic dataset summarizing essential information of a large-scale dataset and thus can be used as a proxy dataset for scalable Bayesian inference. Typically, a Bayesian pseudocoreset is constructed by minimizing a divergence measure between the posterior conditioning on the pseudocoreset and the posterior conditioning on the full dataset. However, evaluating the divergence can be challenging, particularly for the models like deep neural networks having high-dimensional parameters.

We already report the t-SNE visualization of ByPE-V AE and standard V AE in Figure. Figure 6: t-SNE visualization of learned latent representations, colored by labels. Second, we give more generated samples in Fig.8, among Figure 7: Random samples drawn from ByPE-V AEs trained on different datasets. Figure 8: Samples generated by ByPE-V AE based on the same pseudodata point in each plate. In section 5.2, We only report the KNN results of MNIST and Fashion MNIST in the Figure 1.

Recent studies show that advanced priors play a major role in deep generative models. Exemplar V AE, as a variant of V AE with an exemplar-based prior, has achieved impressive results.