A Primal-Dual Message-Passing Algorithm for Approximated Large Scale Structured Prediction

By Tamir Hazan, Raquel Urtasun. 2010

84 citations recorded by the source corpus.

Subfields: Information Extraction, Mathematical Optimization

How researchers see this paper

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

In Information Extraction

In Mathematical Optimization

Citations in the corpus

On Structured Prediction Theory with Calibrated Convex Surrogate Losses (2017, neutral): “, in conditional random fields [25], Φlog(f ,y) := log( ∑ ŷ∈Y exp fŷ)− fy, and their hybrids [38, 21, 22, 41].”
Analyzing Semantic Segmentation Using Hybrid Human-Machine CRFs (2013, background): “The parameters corresponding to different potential terms in the model are learnt in a discriminative fashion [15].”
A High Performance CRF Model for Clothes Parsing (2014, methodology): “To learn the weights, we use the primal-dual method of [30] (we use the implementation of [31]), shown to be more efficient than other structure prediction learning algorithms.”
Stability and Generalization in Structured Prediction (2016, background): “Thus, researchers (Gimpel and Smith, 2010; Hazan and Urtasun, 2010) have proposed a smooth alternative, based on the soft-max function.”
Hierarchical Multitask Structured Output Learning for Large-scale Sequence Segmentation (2011, background): “…node n, one solves the following optimization problem, (w∗n, b ∗ n) = argmin w,b 12 ((1− γ) ||w||2 + γ ∣∣∣∣w−w∗p∣∣∣∣2)+ C ∑ (x,y)∈S ` (〈x,w〉+ b, y) , (1) where p is the parent node of n (with the special case of w∗p = 0 for the root node), ` is an appropriate loss function (e.g., the hinge-loss).”
Omnidirectional 3D reconstruction in augmented Manhattan worlds (2014, methodology): “As (4) depends nonlinearly on τu and τp, traditional CRF learning algorithms [42] are not feasible and we resort to Bayesian optimization [43] for estimating the parameters, yielding wu1 = 1.”
PAC-Bayesian Collective Stability (2014, unknown): “This smoothing between marginal and MAP inference has been explored by a number of authors (e.g., Hazan and Urtasun, 2010; Meshi et al., 2012).”
Holistic Scene Understanding for 3D Object Detection with RGBD Cameras (2013, methodology): “We learn the weights by employing the primal dual learning framework of [11].”
A multi-plane block-coordinate frank-wolfe algorithm for training structural SVMs with a costly max-oracle (2014, background): “This is a question of high practical relevance, since training structured SVMs is known to be computationally costly, especially in a computer vision context where the output set, Y , is large and the max-oracle requires solving a combinatoric optimization problem [23].”
Describing the scene as a whole: Joint object detection, scene classification and semantic segmentation (2012, methodology): “We take advantage of the structure prediction framework of [8] to learn the importance of each of these potentials.”
Deep Structured Prediction with Nonlinear Output Transformations (2018, background): “Several works studied the learning problem when inference is hard [13, 26, 40, 17, 33, 23, 43, 35], designing effective approximations.”
Efficient Exact Inference for 3D Indoor Scene Understanding (2012, unknown): “…ˆ y given an image x we need to solve the following inference task: The model parameters w commonly referred to as weights can be obtained with structured prediction learning algorithms such as structured SVMs [27, 28], conditional random ﬁelds [29] or approximate structured prediction [30].”
Exploring Compositional High Order Pattern Potentials for Structured Output Learning (2013, unknown): “[6, 27] used similar techniques to interpolate “sum” and “min”.”
What Are You Talking About? Text-to-Image Coreference (2014, methodology): “For learning the weights for each potential in our MRF, we use the primal-dual method of [11], specifically, we use the implementation of [26].”
A Sentence Is Worth a Thousand Pixels (2013, methodology): “We learn the weights from training data using the structure prediction framework of [11].”