Learning Latent Subspaces in Variational Autoencoders

Abstract

Variational autoencoders (VAEs) are widely used deep generative models capable of learning unsupervised latent representations of data. Such representations are often difficult to interpret or control. We consider the problem of unsupervised learning of features correlated to specific labels in a dataset. We propose a VAE-based generative model which we show is capable of extracting features correlated to binary labels in the data and structuring it in a latent subspace which is easy to interpret. Our model, the Conditional Subspace VAE (CSVAE), uses mutual information minimization to learn a low-dimensional latent subspace associated with each label that can easily be inspected and independently manipulated. We demonstrate the utility of the learned representations for attribute manipulation tasks on both the Toronto Face and CelebA datasets.

Published as a conference paper at NeurIPS 2018. 15 pages

Cited by in corpus (34)

• Are Disentangled Representations Helpful for Abstract Visual Reasoning?
• Review: Deep Learning in Electron Microscopy
• Weakly Supervised Disentanglement with Guarantees
• Lifelong Teacher-Student Network Learning
• Disentangling Factors of Variation Using Few Labels
• 3DMolNet: A Generative Network for Molecular Structures
• Promises and Pitfalls of Black-Box Concept Learning Models
• Study of Deep Generative Models for Inorganic Chemical Compositions
• An Identifiable Double VAE For Disentangled Representations
• Multi-Agent Reinforcement Learning with Multi-Step Generative Models
• PROUD: PaRetO-gUided Diffusion Model for Multi-objective Generation
• On the Fairness of Disentangled Representations
• MagGAN: High-Resolution Face Attribute Editing with Mask-Guided Generative Adversarial Network
• An Improved Semi-Supervised VAE for Learning Disentangled Representations
• Failure Modes of Variational Autoencoders and Their Effects on Downstream Tasks
• C3VQG: Category Consistent Cyclic Visual Question Generation
• FaceController: Controllable Attribute Editing for Face in the Wild
• Model-Based Robust Deep Learning: Generalizing to Natural, Out-of-Distribution Data
• DualDis: Dual-Branch Disentangling with Adversarial Learning
• Disentangling representations of retinal images with generative models
• Adversarial Disentanglement with Grouped Observations
• Invariance-based Multi-Clustering of Latent Space Embeddings for Equivariant Learning
• Deep Image Synthesis from Intuitive User Input: A Review and Perspectives
• Improving Variational Autoencoder for Text Modelling with Timestep-Wise Regularisation
• Learning Conditional Invariance through Cycle Consistency
• Semi-Supervised Disentanglement of Class-Related and Class-Independent Factors in VAE
• Inverse Learning of Symmetries
• Disentangling the Spatial Structure and Style in Conditional VAE
• Learning Calibratable Policies using Programmatic Style-Consistency
• Group-disentangled Representation Learning with Weakly-Supervised Regularization
• High Mutual Information in Representation Learning with Symmetric Variational Inference
• Proceedings of the First Workshop on Weakly Supervised Learning (WeaSuL)
• Viscos Flows: Variational Schur Conditional Sampling With Normalizing Flows
• TzK: Flow-Based Conditional Generative Model