Deep Value Networks Learn to Evaluate and Iteratively Refine Structured Outputs
arXiv:1703.04363
Abstract
We approach structured output prediction by optimizing a deep value network (DVN) to precisely estimate the task loss on different output configurations for a given input. Once the model is trained, we perform inference by gradient descent on the continuous relaxations of the output variables to find outputs with promising scores from the value network. When applied to image segmentation, the value network takes an image and a segmentation mask as inputs and predicts a scalar estimating the intersection over union between the input and ground truth masks. For multi-label classification, the DVN's objective is to correctly predict the F1 score for any potential label configuration. The DVN framework achieves the state-of-the-art results on multi-label prediction and image segmentation benchmarks.
arXiv commentPublished at ICML 2017
Cited by in corpus (15)
- Flow-based Spatio-Temporal Structured Prediction of Motion Dynamics
- Fast Task-Aware Architecture Inference
- Dynamic Scale Inference by Entropy Minimization
- Structured Output Learning with Conditional Generative Flows
- Energy-Based Models for Continual Learning
- Optimizing Through Learned Errors for Accurate Sports Field Registration
- Neural Message Passing for Multi-Label Classification
- Document-level Event Extraction with Efficient End-to-end Learning of Cross-event Dependencies
- Sketch In, Sketch Out: Accelerating both Learning and Inference for Structured Prediction with Kernels
- Differentiable Forward and Backward Fixed-Point Iteration Layers
- Learning Discriminators as Energy Networks in Adversarial Learning
- Structured Convolutional Kernel Networks for Airline Crew Scheduling
- Structured Prediction in NLP -- A survey
- Search-Guided, Lightly-supervised Training of Structured Prediction Energy Networks
- Adversarial Structure Matching for Structured Prediction Tasks