mlsanityData to 7 Oct 2026

Generalization Gap in Amortized Inference

arXiv 2205.11640 ↗First posted 23 May 2022stat.ML cs.LGcited by 10

Not on the map. Why?

Abstract (arXiv)

The ability of likelihood-based probabilistic models to generalize to unseen data is central to many machine learning applications such as lossless compression. In this work, we study the generalization of a popular class of probabilistic model - the Variational Auto-Encoder (VAE). We discuss the two generalization gaps that affect VAEs and show that overfitting is usually dominated by amortized inference. Based on this observation, we propose a new training objective that improves the generalization of amortized inference. We demonstrate how our method can improve performance in the context of image modeling and lossless compression.

Key facts. Cited by 10 papers in the map (data through 7 Oct 2026). Builds on 30 of its 59 references that are in the map. First posted on arXiv 23 May 2022.

Builds on

30 of 59 references.

References · 30

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  3. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 11,979
  4. 2022Generative Adversarial Networkscited by 5,603
  5. 2017Neural Discrete Representation Learningcited by 4,650
  6. 2013Auto-Encoding Variational Bayescited by 3,980
  7. 2015Variational Inference with Normalizing Flowscited by 2,161
  8. 2016Variational Inference: A Review for Statisticianscited by 1,466
  9. 2018Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representationscited by 929
  10. 2016Pixel Recurrent Neural Networkscited by 814
  11. 2015Importance Weighted Autoencoderscited by 454
  12. 2017PixelCNN++: Improving the PixelCNN with Discretized Logistic Mixture Likelihood and Other Modificationscited by 415
18 more
  1. 2015A note on the evaluation of generative modelscited by 291
  2. 2019Diagnosing and Enhancing VAE Modelscited by 192
  3. 2018Inference Suboptimality in Variational Autoencoderscited by 131
  4. 2017InfoVAE: Information Maximizing Variational Autoencoderscited by 124
  5. 2019Practical Lossless Compression with Latent Variables using Bits Back Codingcited by 85
  6. 2018Bias and Generalization in Deep Generative Models: An Empirical Studycited by 83
  7. 2020Improving Inference for Neural Image Compressioncited by 81
  8. 2016Auxiliary Deep Generative Modelscited by 72
  9. 2017Frequentist Consistency of Variational Bayescited by 60
  10. 2019Bit-Swap: Recursive Bits-Back Coding for Lossless Compression with Hierarchical Latent Variablescited by 51
  11. 2019HiLLoC: Lossless Image Compression with Hierarchical Latent Variable Modelscited by 36
  12. 2021On the Out-of-distribution Generalization of Probabilistic Image Modellingcited by 32
  13. 2014Stochastic Backpropagation and Approximate Inference in Deep Generative Modelscited by 30
  14. 2018Amortized Inference Regularizationcited by 30
  15. 2020Autoencoding Variational Autoencodercited by 24
  16. 2018Spread Divergencecited by 23
  17. 2021Improving Lossless Compression Rates via Monte Carlo Bits-Back Codingcited by 19
  18. 2022Parallel Neural Local Lossless Compressioncited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 3; 2023: 4; 2024: 2; 2025: 0; 2026: 1.

10 papers.

  1. 2023Computationally-Efficient Neural Image Compression with Shallow Decoderscited by 22
  2. 2022Amortized Variational Inference: A Systematic Reviewcited by 16
  3. 2022Laplacian Autoencoders for Learning Stochastic Representationscited by 10
  4. 2023Moment Matching Denoising Gibbs Samplingcited by 7
  5. 2022Improving VAE-based Representation Learningcited by 6
  6. 2026Stop Probing, Start Coding: Why Linear Probes and Sparse Autoencoders Fail at Compositional Generalisationcited by 5
  7. 2023Conditional Sampling of Variational Autoencoders via Iterated Approximate Ancestral Samplingcited by 2
  8. 2024Improving Variational Autoencoder Estimation from Incomplete Data with Mixture Variational Familiescited by 2
  9. 2023A Note on Generalization in Variational Autoencoders: How Effective Is Synthetic Data & Overparameterization?cited by 1
  10. 2024Amortized Variational Inference for Deep Gaussian Processescited by 1