mlsanityData to 7 Oct 2026

Mind the Gap when Conditioning Amortised Inference in Sequential Latent-Variable Models

arXiv 2101.07046 ↗First posted 18 Jan 2021cs.LG stat.MLcited by 14

Not on the map. Why?

Abstract (arXiv)

Amortised inference enables scalable learning of sequential latent-variable models (LVMs) with the evidence lower bound (ELBO). In this setting, variational posteriors are often only partially conditioned. While the true posteriors depend, e.g., on the entire sequence of observations, approximate posteriors are only informed by past observations. This mimics the Bayesian filter -- a mixture of smoothing posteriors. Yet, we show that the ELBO objective forces partially-conditioned amortised posteriors to approximate products of smoothing posteriors instead. Consequently, the learned generative model is compromised. We demonstrate these theoretical findings in three scenarios: traffic flow, handwritten digits, and aerial vehicle dynamics. Using fully-conditioned approximate posteriors, performance improves in terms of generative modelling and multi-step prediction.

Key facts. Cited by 14 papers in the map (data through 7 Oct 2026). Builds on 41 of its 67 references that are in the map. First posted on arXiv 18 Jan 2021.

Builds on

41 of 67 references.

References · 41

  1. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  2. 2013Auto-Encoding Variational Bayescited by 3,980
  3. 2015Variational Inference with Normalizing Flowscited by 2,161
  4. 2016Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Networkcited by 1,898
  5. 2016OpenAI Gymcited by 1,684
  6. 2018Learning Latent Dynamics for Planning from Pixelscited by 1,221
  7. 2015Importance Weighted Autoencoderscited by 454
  8. 2015A Recurrent Latent Variable Model for Sequential Datacited by 362
  9. 2013Black Box Variational Inferencecited by 317
  10. 2018Stochastic Video Generation with a Learned Priorcited by 244
  11. 2019Stochastic Latent Actor-Critic: Deep Reinforcement Learning with a Latent Variable Modelcited by 228
  12. 2016Structured Inference Networks for Nonlinear State Space Modelscited by 182
29 more
  1. 2015Deep Kalman Filterscited by 163
  2. 2016Deep Variational Bayes Filters: Unsupervised Learning of State Space Models from Raw Datacited by 154
  3. 2018Traffic Graph Convolutional Recurrent Neural Network: A Deep Learning Framework for Network-Scale Traffic Learning and Forecastingcited by 145
  4. 2018Inference Suboptimality in Variational Autoencoderscited by 131
  5. 2018Sequential Attend, Infer, Repeat: Generative Modelling of Moving Objectscited by 128
  6. 2018Learning to Decompose and Disentangle Representations for Video Predictioncited by 120
  7. 2016Sequential Neural Models with Stochastic Layerscited by 115
  8. 2017Adversarial Variational Bayes: Unifying Variational Autoencoders and Generative Adversarial Networkscited by 107
  9. 2017A Disentangled Recognition and Nonlinear Dynamics Model for Unsupervised Learningcited by 107
  10. 2016R'enyi Divergence Variational Inferencecited by 104
  11. 2017Filtering Variational Objectivescited by 91
  12. 2016Variational inference for Monte Carlo objectivescited by 83
  13. 2018Variational Autoencoder with Arbitrary Conditioningcited by 68
  14. 2018Deep Bidirectional and Unidirectional LSTM Recurrent Neural Network for Network-wide Traffic Speed Predictioncited by 61
  15. 2014Learning Stochastic Recurrent Networkscited by 60
  16. 2014Variational Recurrent Auto-Encoderscited by 53
  17. 2019Learning Hierarchical Priors in VAEscited by 53
  18. 2018Learning and Querying Fast Generative Models for Reinforcement Learningcited by 48
  19. 2018Temporal Difference Variational Auto-Encodercited by 48
  20. 2015Black box variational inference for state space modelscited by 31
  21. 2020Learning to Fly via Deep Model-Based Reinforcement Learningcited by 19
  22. 2017Unsupervised Real-Time Control through Variational Empowermentcited by 17
  23. 2019Switching Linear Dynamics for Variational Bayes Filteringcited by 17
  24. 2016Variational Inference for On-line Anomaly Detection in High-Dimensional Time Seriescited by 12
  25. 2018Approximate Bayesian inference in spatial environmentscited by 8
  26. 2018The Blackbird Dataset: A large-scale dataset for UAV perception in aggressive flightcited by 8
  27. 2019Note on the bias and variance of variational inferencecited by 4
  28. 2019Re-examination of the Role of Latent Variables in Sequence Modelingcited by 2
  29. 2019Variational Tracking and Prediction with Generative Disentangled State-Space Modelscited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 7; 2023: 5; 2024: 0; 2025: 0; 2026: 2.

14 papers.

Sort:

  1. 2022Unsupervised Time-Series Representation Learning with Iterative Bilinear Temporal-Spectral Fusioncited by 96
  2. 2023TimesURL: Self-supervised Contrastive Learning for Universal Time Series Representation Learningcited by 51
  3. 2022Tractable Dendritic RNNs for Reconstructing Nonlinear Dynamical Systemscited by 34
  4. 2022Integrating Multimodal Data for Joint Generative Modeling of Complex Dynamicscited by 13
  5. 2022Bounding Evidence and Estimating Log-Likelihood in VAEcited by 6
  6. 2023Cheap and Deterministic Inference for Deep State-Space Models of Interacting Dynamical Systemscited by 5
  7. 2023Laser: Latent Set Representations for 3D Generative Modelingcited by 3
  8. 2023Sampling-Free Probabilistic Deep State-Space Modelscited by 2
  9. 2026Latent Matters: Learning Deep State-Space Modelscited by 2
  10. 2022Benchmarking Generative Latent Variable Models for Speechcited by 1
  11. 2023On the Natural Gradient of the Evidence Lower Boundcited by 1
  12. 2026The Dynamic-Probabilistic Consistency Gap in Chaotic Surrogate Modelingcited by 1
2 more
  1. 2022Linear Variational State-Space Filteringcited by 0
  2. 2022Amortized backward variational inference in nonlinear state-space modelscited by 0