mlsanityData to 14 Oct 2026

Implicit Training of Energy Model for Structure Prediction

arXiv 2211.11649 ↗First posted 21 Nov 2022cs.LG cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Most deep learning research has focused on developing new model and training procedures. On the other hand the training objective has usually been restricted to combinations of standard losses. When the objective aligns well with the evaluation metric, this is not a major issue. However when dealing with complex structured outputs, the ideal objective can be hard to optimize and the efficacy of usual objectives as a proxy for the true objective can be questionable. In this work, we argue that the existing inference network based structure prediction methods ( Tu and Gimpel 2018; Tu, Pang, and Gimpel 2020) are indirectly learning to optimize a dynamic loss objective parameterized by the energy model. We then explore using implicit-gradient based technique to learn the corresponding dynamic objectives. Our experiments show that implicitly learning a dynamic loss landscape is an effective method for improving model performance in structure prediction.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 42 of its 82 references that are in the map. First posted on arXiv 21 Nov 2022.

Builds on

42 of 82 references.

References · 42

  1. 2017Attention Is All You Needcited by 60,361
  2. 2015Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networkscited by 9,639
  3. 2016Improved Techniques for Training GANscited by 2,970
  4. 2016Identity Mappings in Deep Residual Networkscited by 1,965
  5. 2016Learning to learn by gradient descent by gradient descentcited by 620
  6. 2017OptNet: Differentiable Optimization as a Layer in Neural Networkscited by 599
  7. 2019Meta-Learning with Implicit Gradientscited by 525
  8. 2019Differentiable Convex Optimization Layerscited by 465
  9. 2019Meta-Learning with Differentiable Convex Optimizationcited by 461
  10. 2018Bilevel Programming for Hyperparameter Optimization and Meta-Learningcited by 405
  11. 2015Gradient-based Hyperparameter Optimization through Reversible Learningcited by 360
  12. 2019Optimizing Millions of Hyperparameters by Implicit Differentiationcited by 304
30 more
  1. 2017Forward and Reverse Gradient-Based Hyperparameter Optimizationcited by 234
  2. 2019Differentiation of Blackbox Combinatorial Solverscited by 226
  3. 2016Learning to Optimizecited by 195
  4. 2018Meta-Reinforcement Learning of Structured Exploration Strategiescited by 130
  5. 2018Noise Contrastive Estimation and Negative Sampling for Conditional Models: Consistency and Statistical Efficiencycited by 105
  6. 2018On Learning Intrinsic Rewards for Policy Gradient Methodscited by 102
  7. 2020Residual Energy-Based Models for Text Generationcited by 95
  8. 2018SGM: Sequence Generation Model for Multi-label Classificationcited by 89
  9. 2018Evolved Policy Gradientscited by 88
  10. 2020Discovering Reinforcement Learning Algorithmscited by 77
  11. 2019Meta-Learning via Learned Losscited by 60
  12. 2015Structured Prediction Energy Networkscited by 59
  13. 2017Classical Structured Prediction Losses for Sequence to Sequence Learningcited by 55
  14. 2020Energy-Based Reranking: Improving Neural Machine Translation Using Energy-Based Modelscited by 44
  15. 2018Learning to Teach with Dynamic Loss Functionscited by 40
  16. 2019Structured Output Learning with Conditional Generative Flowscited by 37
  17. 2020ENGINE: Energy-Based Inference Networks for Non-Autoregressive Machine Translationcited by 34
  18. 2017End-to-End Learning for Structured Prediction Energy Networkscited by 33
  19. 2019Addressing the Loss-Metric Mismatch with Adaptive Loss Alignmentcited by 33
  20. 2017Learning to Learn: Meta-Critic Networks for Sample Efficient Learningcited by 28
  21. 2019Guided Meta-Policy Searchcited by 26
  22. 2015Training Deep Neural Networks via Direct Loss Minimizationcited by 25
  23. 2021Enhancing Label Correlation Feedback in Multi-Label Text Classification via Multi-Task Learningcited by 15
  24. 2018Learning Approximate Inference Networks for Structured Predictioncited by 13
  25. 2017Deep Value Networks Learn to Evaluate and Iteratively Refine Structured Outputscited by 12
  26. 2019Order-free Learning Alleviating Exposure Bias in Multi-label Classificationcited by 12
  27. 2019Improving Joint Training of Inference Networks and Structured Prediction Energy Networkscited by 6
  28. 2017Online Learning of a Memory for Learning Ratescited by 4
  29. 2018Search-Guided, Lightly-supervised Training of Structured Prediction Energy Networkscited by 4
  30. 2019Benchmarking Approximate Inference Methods for Neural Structured Predictioncited by 3

Built on by

Not cited yet.