HQ-VAE: Hierarchical Discrete Representation Learning with Variational Bayes

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Takida, Yuhta, Ikemiya, Yukara, Shibuya, Takashi, Shimada, Kazuki, Choi, Woosung, Lai, Chieh-Hsin, Murata, Naoki, Uesaka, Toshimitsu, Uchida, Kengo, Liao, Wei-Hsiang, Mitsufuji, Yuki
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917624545280000
author Takida, Yuhta
Ikemiya, Yukara
Shibuya, Takashi
Shimada, Kazuki
Choi, Woosung
Lai, Chieh-Hsin
Murata, Naoki
Uesaka, Toshimitsu
Uchida, Kengo
Liao, Wei-Hsiang
Mitsufuji, Yuki
author_facet Takida, Yuhta
Ikemiya, Yukara
Shibuya, Takashi
Shimada, Kazuki
Choi, Woosung
Lai, Chieh-Hsin
Murata, Naoki
Uesaka, Toshimitsu
Uchida, Kengo
Liao, Wei-Hsiang
Mitsufuji, Yuki
contents Vector quantization (VQ) is a technique to deterministically learn features with discrete codebook representations. It is commonly performed with a variational autoencoding model, VQ-VAE, which can be further extended to hierarchical structures for making high-fidelity reconstructions. However, such hierarchical extensions of VQ-VAE often suffer from the codebook/layer collapse issue, where the codebook is not efficiently used to express the data, and hence degrades reconstruction accuracy. To mitigate this problem, we propose a novel unified framework to stochastically learn hierarchical discrete representation on the basis of the variational Bayes framework, called hierarchically quantized variational autoencoder (HQ-VAE). HQ-VAE naturally generalizes the hierarchical variants of VQ-VAE, such as VQ-VAE-2 and residual-quantized VAE (RQ-VAE), and provides them with a Bayesian training scheme. Our comprehensive experiments on image datasets show that HQ-VAE enhances codebook usage and improves reconstruction performance. We also validated HQ-VAE in terms of its applicability to a different modality with an audio dataset.
format Preprint
id arxiv_https___arxiv_org_abs_2401_00365
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle HQ-VAE: Hierarchical Discrete Representation Learning with Variational Bayes
Takida, Yuhta
Ikemiya, Yukara
Shibuya, Takashi
Shimada, Kazuki
Choi, Woosung
Lai, Chieh-Hsin
Murata, Naoki
Uesaka, Toshimitsu
Uchida, Kengo
Liao, Wei-Hsiang
Mitsufuji, Yuki
Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
Vector quantization (VQ) is a technique to deterministically learn features with discrete codebook representations. It is commonly performed with a variational autoencoding model, VQ-VAE, which can be further extended to hierarchical structures for making high-fidelity reconstructions. However, such hierarchical extensions of VQ-VAE often suffer from the codebook/layer collapse issue, where the codebook is not efficiently used to express the data, and hence degrades reconstruction accuracy. To mitigate this problem, we propose a novel unified framework to stochastically learn hierarchical discrete representation on the basis of the variational Bayes framework, called hierarchically quantized variational autoencoder (HQ-VAE). HQ-VAE naturally generalizes the hierarchical variants of VQ-VAE, such as VQ-VAE-2 and residual-quantized VAE (RQ-VAE), and provides them with a Bayesian training scheme. Our comprehensive experiments on image datasets show that HQ-VAE enhances codebook usage and improves reconstruction performance. We also validated HQ-VAE in terms of its applicability to a different modality with an audio dataset.
title HQ-VAE: Hierarchical Discrete Representation Learning with Variational Bayes
topic Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2401.00365