Bayes-Factor-VAE: Hierarchical Bayesian Deep Auto-Encoder Models for Factor Disentanglement

We propose a family of novel hierarchical Bayesian deep auto-encoder models capable of identifying disentangled factors of variability in data. While many recent attempts at factor disentanglement have focused on sophisticated learning objectives within the VAE framework, their choice of a standard normal as the latent factor prior is both suboptimal and detrimental to performance. Our key observation is that the disentangled latent variables responsible for major sources of variability, the relevant factors, can be more appropriately modeled using long-tail distributions. The typical Gaussian priors are, on the other hand, better suited for modeling of nuisance factors. Motivated by this, we extend the VAE to a hierarchical Bayesian model by introducing hyper-priors on the variances of Gaussian latent priors, mimicking an infinite mixture, while maintaining tractable learning and inference of the traditional VAEs. This analysis signifies the importance of partitioning and treating in a different manner the latent dimensions corresponding to relevant factors and nuisances. Our proposed models, dubbed Bayes-Factor-VAEs, are shown to outperform existing methods both quantitatively and qualitatively in terms of latent disentanglement across several challenging benchmark tasks.

[1]  K. Do,et al.  Efficient and Adaptive Estimation for Semiparametric Models. , 1994 .

[2]  Sami Romdhani,et al.  A 3D Face Model for Pose and Illumination Invariant Face Recognition , 2009, 2009 Sixth IEEE International Conference on Advanced Video and Signal Based Surveillance.

[3]  Erkki Oja,et al.  Independent Component Analysis , 2001 .

[4]  Dana H. Brooks,et al.  Structured Disentangled Representations , 2018, AISTATS.

[5]  R. O’Hara,et al.  A review of Bayesian variable selection methods: what, how and which , 2009 .

[6]  Max Welling,et al.  VAE with a VampPrior , 2017, AISTATS.

[7]  Andriy Mnih,et al.  Disentangling by Factorising , 2018, ICML.

[8]  Vladimir Pavlovic,et al.  Relevance Factor VAE: Learning and Identifying Disentangled Factors , 2019, ArXiv.

[9]  David Pfau,et al.  Towards a Definition of Disentangled Representations , 2018, ArXiv.

[10]  Abhishek Kumar,et al.  Variational Inference of Disentangled Latent Concepts from Unlabeled Observations , 2017, ICLR.

[11]  Yoshua Bengio,et al.  Learning Independent Features with Adversarial Nets for Non-linear ICA , 2017, 1710.05050.

[12]  Christopher Burgess,et al.  beta-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework , 2016, ICLR 2016.

[13]  Christopher K. I. Williams,et al.  A Framework for the Quantitative Evaluation of Disentangled Representations , 2018, ICLR.

[14]  Yann LeCun,et al.  Disentangling factors of variation in deep representation using adversarial training , 2016, NIPS.

[15]  Navdeep Jaitly,et al.  Adversarial Autoencoders , 2015, ArXiv.

[16]  Michael I. Jordan,et al.  Kernel independent component analysis , 2003 .

[17]  Pascal Vincent,et al.  Representation Learning: A Review and New Perspectives , 2012, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[18]  Bernhard Schölkopf,et al.  Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations , 2018, ICML.

[19]  Nicola De Cao,et al.  Hyperspherical Variational Auto-Encoders , 2018, UAI 2018.

[20]  Roger B. Grosse,et al.  Isolating Sources of Disentanglement in Variational Autoencoders , 2018, NeurIPS.

[21]  Harold Soh,et al.  Hyperprior Induced Unsupervised Disentanglement of Latent Representations , 2018, AAAI.

[22]  Max Welling,et al.  Auto-Encoding Variational Bayes , 2013, ICLR.

[23]  Xiaogang Wang,et al.  Deep Learning Face Attributes in the Wild , 2014, 2015 IEEE International Conference on Computer Vision (ICCV).

[24]  Pieter Abbeel,et al.  InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets , 2016, NIPS.

[25]  Emilien Dupont,et al.  Joint-VAE: Learning Disentangled Joint Continuous and Discrete Representations , 2018, NeurIPS.