### β-VAE: LEARNING BASIC VISUAL CONCEPTS WITH A CONSTRAINED VARIATIONAL FRAMEWORK

309

views

1

Higgins, I., Matthey, L., Pal, A., Burgess, C., Glorot, X., Botvinick, M., … Deepmind, G. (n.d.). β-VAE: LEARNING BASIC VISUAL CONCEPTS WITH A CONSTRAINED VARIATIONAL FRAMEWORK.

Learning an interpretable factorised representation of the independent data gen-erative factors of the world without supervision is an important precursor for the development of artificial intelligence that is able to learn and reason in the same way that humans do. We introduce β-VAE, a new state-of-the-art framework for automated discovery of interpretable factorised latent representations from raw image data in a completely unsupervised manner. Our approach is a modification of the variational autoencoder (VAE) framework. We introduce an adjustable hy-perparameter β that balances latent channel capacity and independence constraints with reconstruction accuracy. We demonstrate that β-VAE with appropriately tuned β > 1 qualitatively outperforms VAE (β = 1), as well as state of the art unsu-pervised (InfoGAN) and semi-supervised (DC-IGN) approaches to disentangled factor learning on a variety of datasets (celebA, faces and chairs). Furthermore, we devise a protocol to quantitatively compare the degree of disentanglement learnt by different models, and show that our approach also significantly outperforms all baselines quantitatively. Unlike InfoGAN, β-VAE is stable to train, makes few assumptions about the data and relies on tuning a single hyperparameter β, which can be directly optimised through a hyperparameter search using weakly labelled data or through heuristic visual inspection for purely unsupervised data.

Learning an interpretable factorised representation of the independent data gen-erative factors of the world without supervision is an important precursor for the development of artificial intelligence that is able to learn and reason in the same way that humans do. We introduce β-VAE, a new state-of-the-art framework for automated discovery of interpretable factorised latent representations from raw image data in a completely unsupervised manner. Our approach is a modification of the variational autoencoder (VAE) framework. We introduce an adjustable hy-perparameter β that balances latent channel capacity and independence constraints with reconstruction accuracy. We demonstrate that β-VAE with appropriately tuned β > 1 qualitatively outperforms VAE (β = 1), as well as state of the art unsu-pervised (InfoGAN) and semi-supervised (DC-IGN) approaches to disentangled factor learning on a variety of datasets (celebA, faces and chairs). Furthermore, we devise a protocol to quantitatively compare the degree of disentanglement learnt by different models, and show that our approach also significantly outperforms all baselines quantitatively. Unlike InfoGAN, β-VAE is stable to train, makes few assumptions about the data and relies on tuning a single hyperparameter β, which can be directly optimised through a hyperparameter search using weakly labelled data or through heuristic visual inspection for purely unsupervised data.

Community: paperstack

### 1 Answer

0

**Motivation**

Discover disentangled latent representations in a purely unsupervised manner

**Related work**

Chen, X., Duan, Y., Houthooft, R., Schulman, J., Sutskever, I., & Abbeel, P. (2016). InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets. Retrieved from http://arxiv.org/abs/1606.03657

Kulkarni, T. D., Whitney, W., Kohli, P., & Tenenbaum, J. B. (2015). Deep Convolutional Inverse Graphics Network. Retrieved from http://arxiv.org/abs/1503.03167

**Shortcomings of related work**

require a priori knowledge about the number and nature of generative factors

DC-IGN: semi supervised

InfoGAN: training instability, reduce sample diversity, sensitive to choice of prior distribution

VAE: limited disentanglement performance

**Contributions**

extension of the Variational Auto Encoder

single hyperparameter $$\beta$$

reconstruction fidelity vs quality of disentanglement

controls capacity of the latent information channel and independence pressure

$$\beta = 1$$ is the VAE

$$\beta > 1$$ higher pressure on disentanglement

**Shortcomings**

$$\beta$$ still needs to be optimized using weakly labeled data or visual inspection

**Application to Reinforcement Learning**

Higgins, I., Pal, A., Rusu, A., Matthey, L., Burgess, C., Pritzel, A., … Lerchner, A. (n.d.). DARLA: Improving Zero-Shot Transfer in Reinforcement Learning.

written
10 months ago by
Rylan Schaeffer

Please login to add an answer/comment or follow this question.