Structured Learning from Partial Annotations

Structured learning is appropriate when predicting structured outputs such as trees, graphs, or sequences. Most prior work requires the training set to consist of complete trees, graphs or sequences. Specifying such detailed ground truth can be tedious or infeasible for large outputs. Our main contribution is a large margin formulation that makes structured learning from only partially annotated data possible. The resulting optimization problem is non-convex, yet can be efficiently solve by concave-convex procedure (CCCP) with novel speedup strategies. We apply our method to a challenging tracking-by-assignment problem of a variable number of divisible objects. On this benchmark, using only 25% of a full annotation we achieve a performance comparable to a model learned with a full annotation. Finally, we offer a unifying perspective of previous work using the hinge, ramp, or max loss for structured learning, followed by an empirical comparison on their practical performance.

• Publication year

  2012

• Venue

  International Conference on Machine Learning

• Publication date

  2012-06-26

• Fields of study

  Mathematics, Computer Science

• Identifiers
  arXiv 1206.6421
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Generalization bounds and consistency for latent-structural probit and ramp loss

  2011cited by this paper
• Object Detection with Grammar Models

  2011influential reference
• Learning from Partial Labels

  2011influential reference
• Learning from Partially Annotated Sequences

  2011influential reference
• Structured Learning for Cell Tracking

  2011influential reference
• A Discriminative Latent Model of Object Classes and Attributes

  2010cited by this paper
• Learning from Candidate Labeling Sets

  2010cited by this paper
• Latent hierarchical structural learning for object detection

  2010cited by this paper
• More data means less inference: A pseudo-max approach to structured learning

  2010cited by this paper
• Direct Loss Minimization for Structured Prediction

  2010cited by this paper
• Bundle Methods for Regularized Risk Minimization

  2010cited by this paper
• Learning structural SVMs with latent variables

  2009influential reference
• Structured output regression for detection with partial truncation

  2009cited by this paper
• Predicting Structured Data

  2008cited by this paper
• Tighter Bounds for Structured Estimation

  2008cited by this paper
• Predicting Structured Data

  2007cited by this paper
• Transductive support vector machines for structured variables

  2007cited by this paper
• Discriminative unsupervised learning of structured predictors

  2006cited by this paper
• Learning to Model Spatial Dependency: Semi-Supervised Discriminative Random Fields

  2006cited by this paper
• Large Margin Methods for Structured and Interdependent Output Variables

  2005influential reference
• Maximum Margin Semi-Supervised Learning for Structured Variables

  2005cited by this paper
• Statistical Analysis of Some Multi-Category Large Margin Classification Methods

  2004cited by this paper
• The Concave-Convex Procedure

  2003influential reference
• Learning with Multiple Labels

  2002cited by this paper
• Proximity control in bundle methods for convex nondifferentiable minimization

  1990cited by this paper

• Cross-modality supervised image restoration enables nanoscale tracking of synaptic plasticity in living mice

  2022cites this paper
• Consistent and Tractable Algorithm for Markov Network Learning

  2022cites this paper
• Partial Sequence Labeling With Structured Gaussian Processes

  2022influential citation
• Learning Maximum Margin Markov Networks from examples with missing labels

  2021cites this paper
• Overcoming the curse of dimensionality with Laplacian regularization in semi-supervised learning

  2021cites this paper
• Weak Disambiguation for Partial Structured Output Learning

  2020influential citation
• Overcoming the curse of dimensionality with Laplacian regularization in semi-supervised learning

  2020cites this paper
• Partial Or Complete, That’s The Question

  2019cites this paper
• ilastik: interactive machine learning for (bio)image analysis

  2019cites this paper
• Better Modeling of Incomplete Annotations for Named Entity Recognition

  2019cites this paper
• Piecewise Large Margin Learning for Partially Annotated Sequences

  2019cites this paper
• Learning to Recognize Human Activities Using Soft Labels

  2017cites this paper
• Non-Linear Matrix Completion for Social Image Tagging

  2017cites this paper
• Discriminative learning from partially annotated examples

  2016influential citation
• V-shaped interval insensitive loss for ordinal classification

  2016cites this paper
• Computational Analysis of Behavior.

  2016cites this paper
• Learning to Recognize Human Activities using Soft Labels.

  2016cites this paper
• Image annotation with incomplete labelling by modelling image specific structured loss

  2015cites this paper
• Computer vision and machine learning for microscopy image analysis

  2015cites this paper
• Consistency of structured output learning with missing labels

  2015influential citation
• Flexible and Latent Structured Output Learning - Application to Histology

  2015cites this paper
• Regularization-based multitask learning with applications in computational biology

  2015cites this paper
• A generic methodological framework for studying single cell motility in high-throughput time-lapse data

  2015cites this paper
• Interval Insensitive Loss for Ordinal Classification

  2014influential citation
• Advanced Structured Prediction

  2014cites this paper
• An Overview of Research on Electronic Medical Record Oriented Named Entity Recognition and Entity Relation Extraction

  2014cites this paper
• Structured Learning from Cheap Data

  2014cites this paper
• UvA-DARE ( Digital Academic Repository ) Learning to Recognize Human Activities from Soft Labeled Data

  2014cites this paper
• Learning based biological image analysis

  2014cites this paper
• Tracking Indistinguishable Translucent Objects over Time Using Weakly Supervised Structured Learning

  2014cites this paper
• Exploring Image Specific Structured Loss for Image Annotation with Incomplete Labelling

  2014cites this paper
• Multi-utility Learning: Structured-Output Learning with Multiple Annotation-Specific Loss Functions

  2014cites this paper
• Effective Active Learning Strategies for the Use of Large-Margin Classifiers in Semantic Annotation: An Optimal Parameter Discovery Perspective

  2014cites this paper
• Regularization-Based Multitask Learning With Applications to Genome Biology and Biological Imaging

  2014influential citation
• Learning to Recognize Human Activities from Soft Labeled Data

  2014cites this paper
• Statistical formulation of structured output learning from partially annotated examples

  2013influential citation
• Improving Chinese Word Segmentation Using Partially Annotated Sentences

  2013cites this paper
• 1 Structured Learning from Cheap Data

  2013cites this paper
• Discriminative structured output learning from partially annotated examples

  2012cites this paper
• Visual tracking via weakly supervised learning from multiple imperfect oracles

  2010cites this paper
• алгоритма семантической

  year unknowncites this paper