Generalization Bounds and Consistency for Structured Labeling

70 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

On Structured Prediction Theory with Calibrated Convex Surrogate Losses (2017, background): “It is known that Crammer-Singer multi-class SVM [15], which SSVM is built on, is not consistent for 0-1 loss unless there is a majority class with probability at least 12 [49, 31].”
Structured Learning with Approximate Inference (2007, methodology): “We argue that without understanding combinations of inference and learning, such as these, that are appropriately compatible, learning performance under approximate inference cannot be guaranteed.”
Bundle Methods for Regularized Risk Minimization (2010, background): “Extensions of these loss functions allow us to handle structure in the output space (Bakir et al., 2007) (also see Appendix A for a comprehensive exposition of many common loss functions).”
Structured Learning and Prediction in Computer Vision (2011, background): “Optimizing a convex upper bound nevertheless makes sense, see [McAllester, 2007] for an in-depth discussion.”
Multi-output learning via spectral filtering (2012, methodology): “We refer to the papers by Schwartz (1964), Micchelli and Pontil (2005), Carmeli et al.”
A Hybrid Loss for Multiclass and Structured Prediction (2014, unknown): “P (w) = 1 Z exp(− ‖w‖(2) 2 ) and the posterior Q(w′) = 1 Z exp(− ‖w′−w‖2 2 ), one can show eD = PD(EQM(w, y) ≤ 0) by symmetry argument proposed in [14, 17].”