Consistency Regularization for Variational Auto-Encoders

By Samarth Sinha, Adji Bousso Dieng. 2021

80 citations recorded by the source corpus.

Variational auto-encoders (VAEs) are a powerful approach to unsupervised learning. They enable scalable approximate posterior inference in latent-variable models using variational inference (VI). A VAE posits a variational family parameterized by a deep neural network called an encoder that takes data as input. This encoder is shared across all the observations, which amortizes the cost of inference. However the encoder of a VAE has the undesirable property that it maps a given observation and a semantics-preserving transformation of it to different latent representations. This"inconsistency"of the encoder lowers the quality of the learned representations, especially for downstream tasks, and also negatively affects generalization. In this paper, we propose a regularization method to enforce consistency in VAEs. The idea is to minimize the Kullback-Leibler (KL) divergence between the variational distribution when conditioning on the observation and the variational distribution when conditioning on a random semantic-preserving transformation of this observation. This regularization is applicable to any VAE. In our experiments we apply it to four different VAE variants on several benchmark datasets and found it always improves the quality of the learned representations but also leads to better generalization. In particular, when applied to the Nouveau Variational Auto-Encoder (NVAE), our regularization method yields state-of-the-art performance on MNIST and CIFAR-10. We also applied our method to 3D data and found it learns representations of superior quality as measured by accuracy on a downstream classification task.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Generative Models

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Generative Models

Citations in the corpus

Maximum Likelihood Training of Score-Based Diffusion Models (2021, background): “The likelihood for CIFAR-10 can be significantly improved by incorporating advanced data augmentation, as demonstrated in [25, 41].”
Score Matching Model for Unbounded Data Score (2021, methodology): “Additionally, UDM trained on CelebA64 is the best performer in terms of the FID performance, but CR-NVAE [26] outperforms our model in NLL, in which we emphasize that there is a chance of improving NLL with UDM by reducing the dequantization gap [42] with alternative dequantization schemes [39, 38]…”
QARV: Quantization-Aware ResNet VAE for Lossy Image Compression (2023, background): “ResNet VAE [11], a class of hierarchical VAEs, provides the most promising performance in terms of NLL on images [12], [34].”