Adversarial Constraint Learning for Structured Prediction
This addresses the problem of expensive labeling for structured prediction tasks in domains like computer vision and time series analysis, offering a semi-supervised approach that is novel but builds on existing adversarial and constraint-based methods.
The paper tackles the problem of reducing label collection burden in structured prediction by learning constraints from black-box simulators rather than requiring manual specification, achieving high accuracy with few or no labeled inputs across tracking, pose estimation and time series prediction tasks.
Constraint-based learning reduces the burden of collecting labels by having users specify general properties of structured outputs, such as constraints imposed by physical laws. We propose a novel framework for simultaneously learning these constraints and using them for supervision, bypassing the difficulty of using domain expertise to manually specify constraints. Learning requires a black-box simulator of structured outputs, which generates valid labels, but need not model their corresponding inputs or the input-label relationship. At training time, we constrain the model to produce outputs that cannot be distinguished from simulated labels by adversarial training. Providing our framework with a small number of labeled inputs gives rise to a new semi-supervised structured prediction model; we evaluate this model on multiple tasks --- tracking, pose estimation and time series prediction --- and find that it achieves high accuracy with only a small number of labeled inputs. In some cases, no labels are required at all.