Learning Latent Subspaces in Variational Autoencoders

By Jack Klys, Jake Snell, Richard S. Zemel. 2018

156 citations recorded by the source corpus.

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.

Subfields: Computer Vision and Pattern Recognition, General Social Sciences, Generative Models, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In General Social Sciences

In Generative Models

In Representation & Self-Supervised Learning

Citations in the corpus

Flexibly Fair Representation Learning by Disentanglement (2019, background): “Several recent VAE variants have incorporated label information into latent vari-Flexibly able learning for image synthesis (Klys et al., 2018) and single-attribute fair representation learning (Song et al., 2019; Botros & Tomczak, 2018; Moyer et al., 2018).”
Weakly Supervised Disentanglement with Guarantees (2019, background): “Learning such representations that align with the underlying factors of variation may be critical to the development of machine learning models that are explainable or human-controllable (Gilpin et al., 2018; Lee et al., 2019; Klys et al., 2018).”
On the Fairness of Disentangled Representations (2019, background): “[77, 15, 66, 70, 50, 52, 1] assume partially observed factors of variation that should be disentangled from the other unobserved ones.”
Are Disentangled Representations Helpful for Abstract Visual Reasoning? (2019, background): “Due to the initial poor performance of purely unsupervised methods, the field initially focused on semi-supervised [62, 11, 57, 58, 44, 46] and weakly supervised approaches [31, 12, 40, 21, 78, 20, 15, 35, 80, 54, 47, 64, 8].”
Disentangling Factors of Variation Using Few Labels (2019, background): “[34] Jack Klys, Jake Snell, and Richard Zemel.”
Lifelong Teacher-Student Network Learning (2021, methodology): “The loss function used for training hzðtjzÞ and pu1ðzjxÞ is defined as the crossentropy loss L, optimized as, [41], [42], [43] YE AND BORS: LIFELONG TEACHER-STUDENT NETWORK LEARNING 6285”
Disentangling Factors of Variations Using Few Labels (2020, background): “Supervision and inductive biases: While there is prior work on semi-supervised disentanglement learning (Reed et al., 2014; Cheung et al., 2014; Mathieu et al., 2016; Narayanaswamy et al., 2017; Kingma et al., 2014; Klys et al., 2018), these methods aim to disentangle only some observed factors of v”
Learning joint latent representations based on information maximization (2021, methodology): “Similarly to the approach from [15], we propose to use an approximate posterior q / ð y j z Þ , modelled by a variational encoder of parameters / , aiming to approximate p ð y j z Þ , and q d ð z j x Þ which is implemented by a variational encoder of parameters d , for approximating the conditional…”