Improving Variational Autoencoder Estimation from Incomplete Data with Mixture Variational Families
Not on the map. Why?
Abstract (arXiv)
We consider the task of estimating variational autoencoders (VAEs) when the training data is incomplete. We show that missing data increases the complexity of the model's posterior distribution over the latent variables compared to the fully-observed case. The increased complexity may adversely affect the fit of the model due to a mismatch between the variational and model posterior distributions. We introduce two strategies based on (i) finite variational-mixture and (ii) imputation-based variational-mixture distributions to address the increased posterior complexity. Through a comprehensive evaluation of the proposed approaches, we show that variational mixtures are effective at improving the accuracy of VAE estimation from incomplete data.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 36 of its 57 references that are in the map. First posted on arXiv 5 Mar 2024.
Builds on
36 of 57 references.
References · 36
- 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
- 2022Generative Adversarial Networkscited by 5,603
- 2013Auto-Encoding Variational Bayescited by 3,980
- 2016Density estimation using Real NVPcited by 2,029
- 2019Normalizing Flows for Probabilistic Modeling and Inferencecited by 1,315
- 2019On the Convergence of Adam and Beyondcited by 834
- 2016Automatic chemical design using a data-driven continuous representation of moleculescited by 762
- 2017Masked Autoregressive Flow for Density Estimationcited by 680
- 2015Importance Weighted Autoencoderscited by 454
- 2018Multimodal Generative Models for Scalable Weakly-Supervised Learningcited by 200
- 2016Structured Inference Networks for Nonlinear State Space Modelscited by 182
- 2019Variational Mixture-of-Experts Autoencoders for Multi-Modal Deep Generative Modelscited by 177
24 more
- 2018Handling Incomplete Heterogeneous Data using VAEscited by 148
- 2018Inference Suboptimality in Variational Autoencoderscited by 131
- 2017Sticking the Landing: Simple, Lower-Variance Gradient Estimators for Variational Inferencecited by 97
- 2018Implicit Reparameterization Gradientscited by 84
- 2016Variational inference for Monte Carlo objectivescited by 83
- 2018Tighter Variational Bounds are Not Necessarily Bettercited by 76
- 2014Reweighted Wake-Sleepcited by 66
- 2020not-MIWAE: Deep Generative Modelling with Missing not at Random Datacited by 61
- 2017Generative Models of Visually Grounded Imaginationcited by 44
- 2018Doubly Reparameterized Gradient Estimators for Monte Carlo Objectivescited by 41
- 2017Latent Constraints: Learning to Generate Conditionally from Unconditional Generative Modelscited by 39
- 2017Reinterpreting Importance-Weighted Autoencoderscited by 36
- 2021Identifiable Generative Models for Missing Not at Random Data Imputationcited by 30
- 2020f-Divergence Variational Inferencecited by 28
- 2018Leveraging the Exact Likelihood of Deep Latent Variable Modelscited by 23
- 2020Automatic Differentiation Variational Inference with Mixturescited by 19
- 2013What Is Meant by "Missing at Random"?cited by 16
- 2021On Incorporating Inductive Biases into VAEscited by 11
- 2022Generalization Gap in Amortized Inferencecited by 10
- 2018Autoencoders and Probabilistic Inference with Missing Data: An Exact Solution for The Factor Analysis Casecited by 6
- 2022Cooperation in the Latent Space: The Benefits of Adding Mixture Components in Variational Autoencoderscited by 6
- 2021Variational Gibbs Inference for Statistical Model Estimation from Incomplete Datacited by 4
- 2023Conditional Sampling of Variational Autoencoders via Iterated Approximate Ancestral Samplingcited by 2
- 2023Posterior Consistency for Missing Data in Variational Autoencoderscited by 1
Built on by
Citing papers per year: 2024: 1; 2025: 1.
2 papers.