BIVA: A Very Deep Hierarchy of Latent Variables for Generative Modeling

NeurIPS 2019 Lars MaaløeMarco FraccaroValentin LiévinOle Winther

With the introduction of the variational autoencoder (VAE), probabilistic latent variable models have received renewed attention as powerful generative models. However, their performance in terms of test likelihood and quality of generated samples has been surpassed by autoregressive models without stochastic units... (read more)

PDF Abstract

Results from the Paper


TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT BENCHMARK
Image Generation CIFAR-10 BIVA Maaloe et al. (2019) bits/dimension 3.08 # 14
Image Generation ImageNet 32x32 BIVA Maaloe et al. (2019) bpd 3.96 # 7

Methods used in the Paper