[논문 리뷰] Isolating Sources of Disentanglement in Variational Autoencoders

Ricky T. Q. Chen, Xuechen Li외 2명|arXiv (Cornell University)|2018. 02. 14.

Generative Adversarial Networks and Image Synthesis참고 문헌 46인용 수 145

한 줄 요약

이 논문은 ELBO를 분해하여 총상관 항을 분리하고, 추가 하이퍼파라미터 없이 beta-VAE를 보완하는 plug-in으로 beta-TCVAE를 도입하며, classifier-free disentanglement 지표 MIG를 제안한다. 또한 총상관과 데이터셋 전반의 disentanglement를 경험적으로 연결한다.

ABSTRACT

We decompose the evidence lower bound to show the existence of a term measuring the total correlation between latent variables. We use this to motivate our $β$-TCVAE (Total Correlation Variational Autoencoder), a refinement of the state-of-the-art $β$-VAE objective for learning disentangled representations, requiring no additional hyperparameters during training. We further propose a principled classifier-free measure of disentanglement called the mutual information gap (MIG). We perform extensive quantitative and qualitative experiments, in both restricted and non-restricted settings, and show a strong relation between total correlation and disentanglement, when the latent variables model is trained using our framework.

연구 동기 및 목표

• Motivate and quantify disentanglement in VAEs by decomposing the ELBO to identify the total correlation term.
• Propose a training method that weights decomposition terms without introducing new hyperparameters.
• Introduce beta-TCVAE as a plug-in replacement for beta-VAE with automatic disentanglement benefits.
• Propose a classifier-free, information-theoretic metric (MIG) to evaluate disentanglement across latent distributions.

제안 방법

• Derive an ELBO decomposition revealing index-code MI, total correlation, and dimension-wise KL terms.
• Propose minibatch-weighted sampling to estimate decomposition terms without extra hyperparameters.
• Define beta-TCVAE as a special case with alpha=gamma=1 and beta controlling TC penalty.
• Provide an alternative training approach to estimate TC without a discriminator.

실험 결과

연구 질문

• RQ1 Does penalizing the total correlation term in the ELBO promote disentanglement in VAEs?
• RQ2 Can beta-TCVAE achieve better disentanglement than beta-VAE without adding training complexity?
• RQ3 Is there a robust, classifier-free metric to quantify disentanglement across latent distributions?
• RQ4 How does total correlation correlate with disentanglement across datasets and sampling biases?

주요 결과

• beta-TCVAE yields more interpretable disentangled representations than beta-VAE in several datasets.
• Total correlation correlates negatively with disentanglement under beta-TCVAE, supporting the TC penalty's role.
• MIG provides a classifier-free, axis-aligned, generalizable disentanglement measure applicable to various latent distributions.
• The proposed minibatch weighting allows training with TC weighting without additional hyperparameters.
• FactorVAE, which is similar in objective, can be outperformed when density ratio tricks are hard to train, highlighting beta-TCVAE robustness.
• beta-TCVAE remains effective even under non-uniform or dependent factor sampling, improving interpretability over baselines.]
• table_headers: [],
• table_rows: []} } - This JSON got malformed. Let me fix. The
• inserted incorrectly. I'll produce correct JSON. I'll remove stray quotes and ensure proper fields. Also the title field originally should be