Isolating Sources of Disentanglement in Variational Autoencoders

By T. Chen, Xuechen Li, Roger Baker Grosse, David Kristjanson Duvenaud. 2018

1,044 citations recorded by the source corpus.

Subfields: 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 Generative Models

In Representation & Self-Supervised Learning

Citations in the corpus

A Style-Based Generator Architecture for Generative Adversarial Networks (2018, background): “Unfortunately the metrics recently proposed for quantifying disentanglement [25, 30, 6, 18] require an encoder network that maps input images to latent codes.”
Dual Contradistinctive Generative Autoencoder (2020, neutral): “…± 0.42 2 .98% ± 0.14 β -VAE( β =2) [25] 4.69% ± 0.18 5.26% ± 0.22 5.40% ± 0.33 FactorVAE( γ =5) [36] 6.07% ± 0.05 6.18% ± 0.20 6.35% ± 0.48 β -TCVAE ( α =1, β =5, γ =1) [8] 1.62% ± 0.07 1.24% ± 0.05 1.32% ± 0.09 Guided-VAE [12] 1.85% ± 0.08 1.60% ± 0.08 1.49% ± 0.06 Guided-β -TCVAE [12] 1.47% ± 0.”
Diagnosing and Enhancing VAE Models (2019, background): “It has been argued then that adjustments to the VAE objective that push the total correlation towards zero will favor desirable disentangled representations (Chen et al., 2018; Higgins et al., 2017).”
Learning deep representations by mutual information estimation and maximization (2018, background): “Others still have more directly measured the MI between the labels and the representation (Rifai et al., 2012; Chen et al., 2018), which can also reveal the representation’s degree of entanglement.”
Flexibly Fair Representation Learning by Disentanglement (2019, methodology): “In We find that in general, a higher ratio yields lower MIG, but that the highest MIGs are around log γα = −2, with a slight tailing off for smaller ratios.”
On the Transfer of Disentangled Representations in Realistic Settings (2020, methodology): “The Mutual Information Gap (MIG) (Chen et al., 2018), Modularity (Ridgeway and Mozer, 2018), DCI Disentanglement (Eastwood and Williams, 2018) and SAP score (Kumar et al.”
Towards causal generative scene models via competition of experts (2020, background): “, by training a large encoder-decoder architecture with different forms of regularization (Higgins et al.; Kim & Mnih, 2018; Chen et al., 2018; Locatello et al., 2019).”
Weakly-Supervised Disentanglement Without Compromises (2020, background): “…been a renewed interest in the unsupervised learning of disentangled representations (Higgins et al., 2017a; Burgess et al., 2018; Kim & Mnih, 2018; Chen et al., 2018; Kumar et al., 2018) along with quantitative evaluation (Kim & Mnih, 2018; Eastwood & Williams, 2018; Kumar et al., 2018; Chen et…”
Weakly Supervised Disentanglement with Guarantees (2019, methodology): “This is in contrast to the existing disentanglement deﬁnitions and metrics, which require access to the ground truth factors (Higgins et al., 2017; Kumar et al., 2018; Kim & Mnih, 2018; Chen et al., 2018a; Suter et al., 2018; Ridgeway & Mozer, 2018; Eastwood & Williams, 2018).”
On the Transfer of Inductive Bias from Simulation to the Real World: a New Disentanglement Dataset (2019, background): “, FactorVAE [19] or β-TCVAE [5], while DIP-VAE proposed to encourage factorization of the inferred prior qφ(z) = ∫ qφ(z|x)p(x) dx.”
Variational Autoencoders Pursue PCA Directions (by Accident) (2018, methodology): “Although quantifying disentanglement is nontrivial, several metrics have been proposed [21, 17, 6].”
Counterfactuals uncover the modular structure of deep generative models (2018, background): “While the β-VAE is designed for extrinsic disentanglement, further work has shown that it can prove suboptimal with respect to other approaches (Chen et al., 2018; Locatello et al., 2018) suggesting further work could investigate whether better extrinsic disentanglement could also favor intrinsic di”
Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations (2018, methodology): “However, Chen et al. (2018) shows in Section H.2 of the Appendix that the bias in the mini-batch estimation of the total correlation does not significantly affect the performances of their model even with small batch sizes.”
Robustly Disentangled Causal Mechanisms: Validating Deep Representations for Interventional Robustness (2018, methodology): “Various methods to validate a learned representation for disentanglement based on known ground truth generative factors G have been proposed (e.g. Eastwood & Williams, 2018; Ridgeway & Mozer, 2018; Chen et al., 2018; Kim & Mnih, 2018).”
Disentangling by Factorising (2018, methodology): “Building on a preliminary version of this paper, (Chen et al., 2018) recently proposed a minibatch-based alternative to our density-ratio-trick-based method for estimating the Total Correlation and introduced an information-theoretic disentangling metric.”
Theory and Evaluation Metrics for Learning Disentangled Representations (2019, background): “The most popular ones are Z-diff score [16, 18], SAP [23] and MIG [8].”
Conditional deep surrogate models for stochastic, high-dimensional, and multi-fidelity systems (2019, background): “…studies that aim to further elucidate the capabilities and limitations of such models [31–36], enhance the interpretability of their results [37–39], as well as establish formal connections with well studied topics in mathematics and statistics, including importance sampling [40,41] and…”
Towards a Definition of Disentangled Representations (2018, background): “Most of the recent approaches to disentangled representation learning [39, 15, 47, 13, 51, 7, 28], as well as the majority of the disentanglement metrics [39, 15, 47, 27] assume that this property should be true.”
Equivariant Hamiltonian Flows (2019, methodology): “For example, most of the disentangling models to date use a unit Gaussian prior [9, 14, 3].”