mlsanityData to 7 Oct 2026

Training Discrete Deep Generative Models via Gapped Straight-Through Estimator

arXiv 2206.07235 ↗First posted 15 Jun 2022cs.LG cs.AIcited by 8

Not on the map. Why?

Abstract (arXiv)

While deep generative models have succeeded in image processing, natural language processing, and reinforcement learning, training that involves discrete random variables remains challenging due to the high variance of its gradient estimation process. Monte Carlo is a common solution used in most variance reduction approaches. However, this involves time-consuming resampling and multiple function evaluations. We propose a Gapped Straight-Through (GST) estimator to reduce the variance without incurring resampling overhead. This estimator is inspired by the essential properties of Straight-Through Gumbel-Softmax. We determine these properties and show via an ablation study that they are essential. Experiments demonstrate that the proposed GST estimator enjoys better performance compared to strong baselines on two discrete deep generative modeling tasks, MNIST-VAE and ListOps.

Key facts. Cited by 8 papers in the map (data through 7 Oct 2026). Builds on 28 of its 36 references that are in the map. First posted on arXiv 15 Jun 2022.

Builds on

28 of 36 references.

References · 28

  1. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 8,964
  2. 2021Evaluating Large Language Models Trained on Codecited by 8,053
  3. 2016Categorical Reparameterization with Gumbel-Softmaxcited by 2,708
  4. 2015Variational Inference with Normalizing Flowscited by 2,161
  5. 2013Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computationcited by 2,031
  6. 2016WaveNet: A Generative Model for Raw Audiocited by 1,710
  7. 2016Generative Adversarial Imitation Learningcited by 1,576
  8. 2017Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environmentscited by 1,519
  9. 2016The Concrete Distribution: A Continuous Relaxation of Discrete Random Variablescited by 1,135
  10. 2019An Introduction to Variational Autoencoderscited by 997
  11. 2014Semi-Supervised Learning with Deep Generative Modelscited by 494
  12. 2016From Softmax to Sparsemax: A Sparse Model of Attention and Multi-Label Classificationcited by 337
16 more
  1. 2015Improved Semantic Representations From Tree-Structured Long Short-Term Memory Networkscited by 305
  2. 2018DeepFakes: a New Threat to Face Recognition? Assessment and Detectioncited by 208
  3. 2018ListOps: A Diagnostic Dataset for Latent Tree Learningcited by 128
  4. 2014Neural Variational Inference and Learning in Belief Networkscited by 116
  5. 2021An Introduction to Deep Generative Modelingcited by 104
  6. 2016Hierarchical Multiscale Recurrent Neural Networkscited by 102
  7. 2016Variational inference for Monte Carlo objectivescited by 83
  8. 2017Improved Variational Autoencoders for Text Modeling using Dilated Convolutionscited by 81
  9. 2020Gradient Estimation with Stochastic Softmax Trickscited by 66
  10. 2019Discretizing Continuous Action Space for On-Policy Optimizationcited by 65
  11. 2017Learning to Compose Task-Specific Tree Structurescited by 48
  12. 2020Rao-Blackwellizing the Straight-Through Gumbel-Softmax Gradient Estimatorcited by 37
  13. 2015MuProp: Unbiased Backpropagation for Stochastic Neural Networkscited by 36
  14. 2020DisARM: An Antithetic Gradient Estimator for Binary Latent Variablescited by 27
  15. 2021Coupled Gradient Estimators for Discrete Latent Variablescited by 9
  16. 2021Soft Actor-Critic With Integer Actionscited by 2

Built on by

Cited by year (2026 to 7 Oct)

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

8 papers.

  1. 2023Bridging Discrete and Backpropagation: Straight-Through and Beyondcited by 31
  2. 2025Efficient Deployment of Spiking Neural Networks on SpiNNaker2 for DVS Gesture Recognition Using Neuromorphic Intermediate Representationcited by 6
  3. 2023Revisiting the Gumbel-Softmax in MADDPGcited by 2
  4. 2023Resampling Gradients Vanish in Differentiable Sequential Monte Carlo Samplerscited by 2
  5. 2024VQSynery: Robust Drug Synergy Prediction With Vector Quantization Mechanismcited by 2
  6. 2024SQUAT: Stateful Quantization-Aware Training in Recurrent Spiking Neural Networkscited by 2
  7. 2025Accelerating Fleet Upgrade Decisions with Machine-Learning Enhanced Optimizationcited by 0
  8. 2026Beyond ReinMax: Low-Variance Gradient Estimators for Discrete Latent Variablescited by 0