Multiple Classifier Systems

Name: Multiple Classifier Systems | 10th International Workshop, MCS 2011, Naples, Italy, June 15-17, 2011. Proceedings
Brand: Springer
Price: 53.49 EUR
Availability: OnlineOnly

10th International Workshop, MCS 2011, Naples, Italy, June 15-17, 2011. Proceedings

Carlo Sansone Josef Kittler Fabio Roli(Editor)

Springer (Publisher)

Published on 15. September 2011

XII, 372 pages

E-Book

PDF with digital watermarking

System requirements

978-3-642-21557-5 (ISBN)

€53.49incl. 7% vat

System requirements

for PDF with digital watermarking

E-Book Single Licence

Available for download

Description

More details

Other editions

Content

Title
Preface
Organization
Table of Contents
Invited Papers
Ensemble Methods for Tracking and Segmentation (Abstract)
Ensembles and Multiple Classifiers: A Game-Theoretic View
Aggregating Strategies
Online Ensembles
Multikernel and Multitask Online Learning
Conclusions
References
Classifier Ensembles I
Anomaly Detection Using Ensembles
Introduction
Datasets
Experiments
Evaluation Metrics
Results
Conclusion
References
Learning to Rank with Nonlinear Monotonic Ensemble
Introduction
The Learning to Rank Problem
Our Method: MonoRank
Nonlinear Monotonic Ensemble: Underlying Theory
The Algorithm
Adaptation of Base Rankers for Usage in Monotonic Ensemble
Reweighing Strategies
Monotonic Aggregating Function
Experimental Results
Yahoo! LETOR 2010 Challenge Dataset
OSHUMED LETOR 3.0 Dataset
Experiment Analysis
Conclusion
References
A Bayesian Approach for Combining Ensembles of GP Classifiers
Introduction
Bayesian Networks for Combining Classifiers
System Architecture
Experimental Results
Conclusions and Future Work
References
Classifier Ensembles II
Multiple Classifiers for Graph of Words Embedding
Introduction
Graph of Words Embedding
Motivation
Embedding Procedure
Vocabulary Selection and Dimensionality Reduction
Multiple Classifiers for Graph of Words Embedding
Multiple Classifier Methods
Multiple Classifiers for Graph of Words Embedding
Classifier Selection
Experimental Results
Databases
Experimental Setup
Results on the Test Set
Conclusions
References
A Dynamic Logistic Multiple Classifier System for Online Classification
Introduction
A Dynamic Logistic Model for Combining Classifiers
Choosing the Component Classifiers
Specification of the ?k(·)
General Applicability
Algorithms for Parameter Updating and Classification
The Predictive Approach to Classification
Implementation
Experimental Comparison
Methods
Datasets and Results
Discussion and Conclusions
Bibliography
Ensemble Methods for Reinforcement Learning with Function Approximation
Introduction
Reinforcement Learning with Parameterized State-Value Functions
Ensemble Methods
Combining the State-Values
Experiments and Results
Maze
Tic-Tac-Toe
Conclusion
References
Trees and Forests
GRASP Forest: A New Ensemble Method for Trees
Introduction
Method
Results
Conclusion and Future Lines
References
Ensembles of Decision Trees for Imbalanced Data
Introduction
Methods
Decision Trees
Ensemble Methods
Experiments
Datasets
Settings
Results
Conclusions and Future Work
References
Compact Ensemble Trees for Imbalanced Data
Introduction
Preliminaries
Properties of -Divergence Criterion
Bootstrap Ensemble of -Trees
Concluding Remarks
References
One-Class Classifiers
Pruned Random Subspace Method for One-Class Classifiers
Introduction
Combining One-Class Classifiers
Pruned Random Subspaces for One-Class Classification
Experiments
Conclusions
References
Approximate Convex Hulls Family for One-Class Classification
Introduction
Approximate Convex Hull Functions Family for One-Class Classification
Approximate Robust Hull Family for One-Class Classification
On the number of Projections in ACH/ARH
Validation and Results
Conclusions
References
Multiple Kernels
Generalized Augmentation of Multiple Kernels
Introduction
Overview of Support Vector Machine
Empirical Feature Function of Kernels
Sum Kernel and Augmented Kernel
Generalized Augmentation Kernel in Multiple-Kernel Spaces
Experimental Results
Experiment 1: Data with Varying Local Distributions
Experiment 2: Benchmark Data
Conclusions
References
A Modified Neutral Point Method for Kernel-Based Fusion of Pattern-Recognition Modalities with Incomplete Data Sets
Introduction
Inferring a Modality-Specific Kernel-Based Classifier from an Unbalanced Training Set
Modality-Specific Kernel Functions
A Single Modality-Specific Kernel-Based Classifier Inferred from an Incomplete Training Set
Neutral Points in the Modality-Specific Linear Space of Object Representation
Fusing Pattern-Recognition Modalities at the Training Stage for Incomplete Data
The Principle of Additive Kernel Fusion
Neutral Point Substitution for Missing Representations of Training Objects
Experiments: Biometric-Based Identity Authentication from Incomplete Data
Conclusions
References
Two-Stage Augmented Kernel Matrix for Object Recognition
Introduction
Linear Combination vs Augmented Kernel Matrix
Two-Stage Multiple Kernel Learning
Kernel Grouping
Selection of Training Samples
Experiments and Discussion
Conclusions
References
Multiple Kernel Learning via Distance Metric Learning for Interactive Image Retrieval
Introduction
Related Work
Multiple Kernel Learning
Distance Metric Learning
Multiple Kernel Learning via Distance Metric Learning
MKL via DML: The Global Version
MKL via DML: The Localised Version
Experiments
Datasets
Results
Conclusions
References
Classifier Selection
Dynamic Ensemble Selection for Off-Line Signature Verification
Introduction
A Hybrid System for Off-Line SV
New Strategies for Dynamic Ensemble Selection
Experimental Methodology
Simulation Results
Conclusions
References
Classifier Selection Approaches for Multi-label Problems
Introduction
A Classifier Selection Approach for Multi-label Problems
Performance Measures for Multi-label Classifiers
Criteria Based on the F Measure for Static Multi-label Classifier Selection
Experimental Evaluation
Conclusions
References
Selection Strategies for pAUC-Based Combination of Dichotomizers
Introduction
ROC Analysis and Partial Area Under the ROC Curve
Combination of Two Dichotomizers
Combination of K&2 Dichotomizers
Single Classifier Selection
Pair Selection
Experimental Results
Conclusions
References
Sequential Combination
Sequential Classifier Combination for Pattern Recognition in Wireless Sensor Networks
Introduction
General Approach
Experiments
First Experiment
Second Experiment
Summary, Discussion and Conclusions
References
Multi-class Multi-scale Stacked Sequential Learning
Introduction
Related Work: Multi-scale Stacked Sequential Learning
Multi-class Multi-scale Stacked Sequential Learning
Extending the Base Classifiers
Extending the Neighborhood Function J
Extended Data Set Grouping: A Compression Approach
Experiments and Results
Conclusions
References
ECOC
A Comparison of Random Forest with ECOC-Based Classifiers
Introduction
ECOC Weighted Decoding
Experiments
Discussion and Conclusions
References
Two Stage Reject Rule for ECOC Classification Systems
Introduction
The ECOC Approach
From an External to an Internal Reject Approach
Experiments
Conclusions and Future Works
References
Introducing the Separability Matrix for Error Correcting Output Codes Coding
Introduction
The Separability Matrix
Application of Separability Matrix for Extension Coding
Training the Base Classifiers
Experimental Results
Conclusions
References
Diversity
Improving Accuracy and Speed of Optimum-Path Forest Classifier Using Combination of Disjoint Training Subsets
Introduction
Optimum-Path Forest Classifier (OPF)
Learning
Combination of Classifiers Trained with Disjoint Subsets
Nearest Neighbor Tie-Breaking
Experiments
Data
Settings and Implementation
Results and Discussion
Improvement in Running Time and Multicore Processing
Conclusions
References
Analyzing the Relationship between Diversity and Evidential Fusion Accuracy
Introduction
Representation of Classifier Outputs
Basics of the Dempster-Shafer (DS) Theory of Evidence
Triplet Mass Function and Computation
Diversity Measures
Experimental Evaluation
Experimental Settings
Results in Decreasing Order
Results in Mixed Order
Summary and Future Work
References
Clustering
Classification by Cluster Analysis: A New Meta-Learning Based Approach
Introduction
Related Work
Classification by Cluster Analysis (CBCA)
Clustering Process
Classification Process
Technical Details of CBCA
Experiments
Implementation Details
Results
Discussion
Conclusion and Future Work
References
C$^3$E: A Framework for Combining Ensembles of Classifiers and Clusterers
Introduction
Description of C3E
Related Work
Experimental Evaluation
Conclusions
References
A Latent Variable Pairwise Classification Model of a Clustering Ensemble
Introduction
Ensemble Model
An Upper Bound for Misclassification Probability
Estimating Characteristics of a Clustering Ensemble
Numerical Experiment
References
CLOOSTING: CLustering Data with bOOSTING
Introduction
Standard and Regularised Adaboost
Using Weak Learners That Abstain
Iterative Clustering with Adaboost
``Subtractive'' Weak Learners
Weak Learners That Abstain
Experimental Evaluation
Conclusions
References
Biometrics
A Geometric Approach to Face Detector Combining
Introduction
Models of Face Representation and Localization Accuracy
A Geometric Method of Face Detectors Combining
Experimental Procedure
Results
Discussion and Conclusion
References
Increase the Security of Multibiometric Systems by Incorporating a Spoofing Detection Algorithm in the Fusion Mechanism
Introduction
Our Approach
Score Fusion Rules
Spoofing Detector
Experimental Results
Dataset
Results
Finding the Best Error Rate at Spoof Detection Level
Conclusions and Future Directions
References
Cohort Based Approach to Multiexpert Class Verification
Introduction
Cohort Based Reasoning
Cohort Score Distribution Modelling
Cohort Score Models in Decision Making
Cohort Based Fusion of Multiple Experts
Experimental Support
Conclusion
References
Computer Security
A Modular Architecture for the Analysis of HTTP Payloads Based on Multiple Classifiers
Introduction
State of the Art
A Modular Architecture for the Analysis of HTTP Payloads
Experimental Setup
Experimental Results
Conclusions
References
Incremental Boolean Combination of Classifiers
Introduction
Learn-and-Combine Approach Using Incremental Boolean Combination
Simulation Results
Conclusions
References
Bagging Classifiers for Fighting Poisoning Attacks in Adversarial Classification Tasks
Introduction
Background
Bagging in the Presence of Outliers
Works on Poisoning Attacks
Motivations of This Work
Problem Formulation and Application Scenarios
Experiments
Experimental Results
Conclusions and Future Work
References
Using a Behaviour Knowledge Space Approach for Detecting Unknown IP Traffic Flows
Introduction
Related Work
The BKS Combiner
Software Tools: TIE and WEKA
Experimental Analysis
Data Set and Base Classifiers
Evaluation of the BKS Approach
Conclusion
References
Author Index

System requirements

Save as PDF Copy link into clipboard

Schweitzer Fachinformationen

Multiple Classifier Systems

Description

More details

Other editions

Additional editions

Content

System requirements