PAC-Bayesian compression bounds on the prediction error of learning algorithms for classification

DOI10.1007/s10994-005-0462-7zbMath1101.68561MaRDI QIDQ5916202

Ralf Herbrich, John Shawe-Taylor, Thore Graepel

Publication date: 17 June 2005

Published in: Machine Learning (Search for Journal in Brave)

Full work available at URL: https://doi.org/10.1007/s10994-005-0462-7

zbMATH Keywords

classification; error bounds; sample compression; kernel classifiers; PAC-Bayes

Mathematics Subject Classification ID

68W05: Nonnumerical algorithms

68T05: Learning and adaptive systems in artificial intelligence

68P30: Coding and information theory (compaction, compression, models of communication, encoding schemes, etc.) (aspects in computer science)

Related Items

Active Nearest-Neighbor Learning in Metric Spaces, Unnamed Item, On the proliferation of support vectors in high dimensions*, The role of mutual information in variational classifiers, Learning the set covering machine by bound minimization and margin-sparsity trade-off, On the perceptron's compression, Learning a priori constrained weighted majority votes, On the fusion of threshold classifiers for categorization and dimensionality reduction

Cites Work

• Unnamed Item
• Unnamed Item
• Unnamed Item
• Modeling by shortest data description
• Support-vector networks
• 10.1162/153244303765208368
• On the Generalization Ability of On-Line Learning Algorithms
• 10.1162/15324430152748236
• A PAC-Bayesian margin bound for linear classifiers
• Structural risk minimization over data-dependent hierarchies
• On the role of pattern matching in information theory
• 10.1162/153244302760200650
• Probability Inequalities for Sums of Bounded Random Variables
• Nearest neighbor pattern classification