Machine Learning and Knowledge Discovery in Databases, Part III

Name: Machine Learning and Knowledge Discovery in Databases, Part III | European Conference, ECML PKDD 2010, Athens, Greece, September 5-9, 2011, Proceedings, Part III
Brand: Springer
Price: 53.49 EUR
Availability: OnlineOnly

European Conference, ECML PKDD 2010, Athens, Greece, September 5-9, 2011, Proceedings, Part III

Dimitrios Gunopulos Thomas Hofmann Donato Malerba Michalis Vazirgiannis(Herausgeber*in)

Springer (Verlag)

Erschienen am 6. September 2011

XXII, 663 Seiten

E-Book

PDF mit Wasserzeichen-DRM

Systemvoraussetzungen

978-3-642-23808-6 (ISBN)

53,49 €inkl. 7% MwSt.

Systemvoraussetzungen

für PDF mit Wasserzeichen-DRM

E-Book Einzellizenz

Als Download verfügbar

Beschreibung

Weitere Details

Weitere Ausgaben

Inhalt

Title page
Welcome to ECML PKDD 2011
Preface
Organization
Table of Contents
Regular Papers
Sparse Kernel-SARSA(?) with an Eligibility Trace
Introduction
Preliminaries
Markov Decision Processes
Reproducing Kernel Hilbert Spaces (RKHS)
Kernel-SARSA(?)
Memory-Efficient Kernel-SARSA (?) Based on the Projectron
Separately Kernelizing the Eligibility Trace
Projected Kernel-SARSA(?) Updates
Empirical Evaluation
Problems
Results
Related Work
Conclusion
References
Influence and Passivity in Social Media
Introduction
Related Work
Twitter
Background on Twitter
Dataset
The IP Algorithm
Evaluation
Computations
Influence as a Correlate of Attention
IP Algorithm Adaptability
Case Studies
Discussion
Conclusion
References
Preference Elicitation and Inverse Reinforcement Learning
Introduction
Formalisation of the Problem
The Statistical Model
Estimation
The Basic Model: A Metropolis-Hastings Procedure
The Augmented Model: A Hybrid Gibbs Procedure
Related Work
Preference Elicitation in User Modelling
Inverse Reinforcement Learning
Experiments
Domains
Algorithms, Priors and Parameters
Performance Measure
Results
Discussion
References
A Novel Framework for Locating Software Faults Using Latent Divergences
Introduction
Related Work
Motivation
An Illustrative Example
Probabilistic Divergences
KL-Divergence
a-Divergences
f-Divergences
Bregman-Divergences
Latent Divergence
Latent Divergences and Fault Localization
Algorithm
Experiments
Siemens Test Suite
Results
Conclusions
References
Transfer Learning with Adaptive Regularizers
Introduction
Regularization Adaptation with Transfer Learning
Theoretical Analysis
Estimation Error
Approximation Error
Approximation Error with Sparse and Concentrated Moment Vectors
Discussion
Algorithmic Details
Experiments
Discussion and Conclusion
References
Multimodal Nonlinear Filtering Using Gauss-Hermite Quadrature
Introduction
Methods
Problem Statement
Fitting a Gaussian Mixture to the Posterior
Gauss-Hermite Quadrature
Related Work
Results
Problem Setup
Representation of Multimodal Posterior Distributions
Convergence When Using Active Learning
Discussion
References
Active Supervised Domain Adaptation
Introduction
Alda: Active Learning Domain Adapted
Preliminaries
Initializing the Uncertainty Sampler
Leveraging Domain Divergence
Hybrid Oracle
Online Alda
Experiments
Setup
B-Alda Results
O-Alda Results
Remarks
Related Work
Discussions and Future Work
References
Efficiently Approximating Markov Tree Bagging for High-Dimensional Density Estimation
Introduction
Graphical Probability Density Model Learning
Bagging in the Context of Learning Bayesian Networks
Mixtures of Markov Trees
Baseline Markov Tree Based Learning Algorithms
The Chow-Liu Algorithm for Learning a Markov Tree
Bagging of Chow-Liu Markov Trees
Inertial Search Heuristic
Proposed Algorithms
Improving the Inertial Search Heuristic by Warm Start
Pruned Mixtures of Bagged Chow-Liu Trees
Experiments
Results in Terms of Accuracies
Computing Times
Conclusion
References
Resource-Aware On-line RFID Localization Using Proximity Data
Introduction
Related Work
Resource-Aware RFID Room-Level-Localization
Resource-Aware RFID Application Scenario
Machine Learning for Prediction Using RFID Data
Advanced Room Prediction Using Contacts
Evaluation
Datasets
Setting
Results and Discussion - Part 1: Machine Learning Baseline
Results and Discussion - Part 2: Utilizing Contact Information
Conclusions
References
On the Stratification of Multi-label Data
Introduction
Stratifying Multi-Label Data
Iterative Stratification
Experiments
Setup
Distribution of Labels and Examples
Variance of Estimates
Conclusions and Future Work
References
Learning First-Order Definite Theories via Object-Based Queries
Introduction
Previous Work
Preliminaries
Base Algorithm
Object-Based Queries
Relevant Object Queries
Eliminating Membership Queries
Imperfect Relevance Oracles
Pairing Queries
Summary and Future Work
References
Fast Support Vector Machines for Structural Kernels
Introduction
Preliminaries: Cutting Plane Algorithm with Sampling
Cutting-Plane Algorithm (Primal)
Cutting-Plane Algorithm (Dual)
Fast CPA for Structural Kernels
Compacting Cutting Plane Models Using DAGs
DAG Tree Kernels
Fast Computation of the MVC on Structural Data
Parallelization
Handling Class-Imbalanced Data
Cost-Proportionate Sampling
Theoretical Analysis of the Algorithm
Experiments
Experimental Setup
Data and Models
Results and Analysis
Related Work
Conclusions and Future Work
References
Generalized Agreement Statistics over Fixed Group of Experts
Introduction
Measuring Inter-expert Agreement
Measuring Agreement against a Group of Experts
Analysis
Properties and Behavior
Empirical Results and Discussion
Evaluating Inter-expert Agreement
Evaluating Agreement against a Group of Fixed Experts
Illustration on Real Data
Related Work
Conclusion and Future Work
References
Compact Coding for Hyperplane Classifiers in Heterogeneous Environment
Introduction
Problem Setting and Preliminaries for Encoding
Compact Coding for Hyperplane Classifiers in Heterogeneous Environment
Macro Level: Arrange Related Tasks
Micro Level Evaluation
The Transfer Learning Algorithm
Experiments
Experimental Setting
Experimental Results
Related Work
Conclusion
References
Multi-label Ensemble Learning
Introduction
Related Work
The EnML Method
Measure Criteria
Multi-objective Optimization Solution
Algorithm Framework
Experiments
Experimental Setup
Performance Comparison
Parameter Settings
Conclusion
References
Rule-Based Active Sampling for Learning to Rank
Introduction
Related Work
Selective Sampling Using Association Rules
Learning to Rank Using Association Rules
Rule-Based Active Sampling
Experimental Evaluation
LETOR Datasets
Results
Discussion
Does the Proposed Sampling Technique Work with Other L2R Methods?
How Does the Proposed Method Fare against Supervised LETOR Baselines?
How Does SSARP Compare to Other Active Learning Methods for L2R?
Conclusions
References
Parallel Structural Graph Clustering
Introduction
Background
Notation and Definitions
Problem Definition
Sequential Structural Clustering
Parallel Structural Graph Clustering
Cluster Comparisons
Size Based Exclusion Criterion
Exclusion Criterion Based on Node Feature Vectors
Definition of a Cluster Representative
Experimental Results
Test Environment and Data Sets
Performance Evaluation
Effects of Algorithm Improvements
Comparison to Sequential Structural Clustering
Experiments on Large Graph Data Sets
Conclusion
References
Aspects of Semi-supervised and Active Learning in Conditional Random Fields
Introduction
Semi-supervised Discriminant Estimator
Conditional Random Fields
Semi-supervised Conditional Random Fields
Motivation for Pool-Based Active Learning
Experiments on Artificial and Real Data Sets
Weighted Conditional Random Fields Experiments
Fully Supervised Active Learning Approach (FuSAL)
Active Learning Experiments on Nettalk Phonetisation Task
Active Learning Experiments on CoNLL 2003 Corpus
Related Work
Conclusion
References
Comparing Probabilistic Models for Melodic Sequences
Introduction
Related Work
Preliminaries
Musical Motifs
Variable Length Markov Model
Restricted Boltzmann Machine
Models
Dirichlet-VMM
Time Convolutional RBM
Experiments
Data Processing and Representation
Implementation Details
Learning Musical Features
Prediction Task
Using the Kullback-Leibler Divergence to Compare Statistics
Discussion
References
Fast Projections onto $ _1,q$-Norm Balls for Grouped Feature Selection
Introduction
Algorithm and Theory
Efficiently Computing Projections
Projections via Proximity
Computing the Proximity Operators
Experimental Results
Projection onto the $ _1,8$-Ball
Projection onto $ _1,q$-Balls
Multitask Lasso with $l_1,8$-Constraint
Simulation Results for MTL
MTL Results on Real-World Data
Conclusions
References
Generalized Dictionary Learning for Symmetric Positive Definite Matrices with Application to Nearest Neighbor Retrieval
Introduction
Background and Related Work
Generalized Dictionary Learning
Online GDL Algorithm
Sparse Coding: Computing b
Nearest Neighbors via GDL
Experiments and Results
Methods Compared against GDL
Dictionary Learning
Experimental Setup
Conclusions
References
Network Regression with Predictive Clustering Trees
Introduction
Related Work
Mining Network Data
Building Predictive Clustering Trees
Learning PCTS from Network Data
The Problem
Measures of Network Autocorrelation
PCTs for Network Regression
Empirical Evaluation
Datasets
Experimental Setup
Results and Discussion
Conclusions
References
Learning from Label Proportions by Optimizing Cluster Model Selection
Introduction
Learning from Label Proportions
The Learning Task
Training and Test Error
Best and Worst Case from a Bayesian Perspective
Learning from Label Proportions by Clustering
Optimization Problem
The LLP Algorithm
Labeling Heuristics
Run-Time Analysis
Generating a Prediction Model
Experiments
Prediction Performance Experiments
Prediction Performance Results
Statistical Significance
Run-Time Comparison
Related Work
Conclusions and Future Work
References
The Minimum Code Length for Clustering Using the Gray Code
Introduction
Notation
Minimum Code Length
Minimizing MCL and Clustering
Problem Formulation
COOL Algorithm
G-COOL: COOL with Gray Code
Gray Code Embedding
Theoretical Analysis of G-COOL
Experiments
Methods
Results and Discussion
Conclusion
References
Learning to Infer Social Ties in Large Networks
Introduction
Problem Definition
Partially-Labeled Pairwise Factor Graph Model (PLP-FGM)
Basic Idea
Partially-Labeled Pairwise Factor Graph Model (PLP-FGM)
Distributed Learning
Experimental Results
Experiment Setup
Accuracy Performance
Scalability Performance
Related Work
Conclusion
References
Comparing Apples and Oranges
Introduction
Preliminaries
Building Global Models from Tiles
Comparing Sets of Tiles
Redescribing Sets of Tiles
Related Work
Experiments
Methods and Mining Results
Measuring Distances
Distances between Results
Redescribing Results
Discussion
Conclusion
References
Learning Monotone Nonlinear Models Using the Choquet Integral
Introduction
Related Work
The Discrete Choquet Integral
Non-additive Measures
Importance of Criteria and Interaction
The Choquet Integral
Choquistic Regression
The Choquistic Model
Parameter Estimation
Experiments
Data Sets
Methods
Results
Concluding Remarks
References
Feature Selection for Transfer Learning
Introduction
Feature Selection with MMD (f-MMD)
Experiments
Synthetic Datasets
Real World Datasets
Related Work
Conclusion
References
Multiview Semi-supervised Learning for Ranking Multilingual Documents
Introduction
Related Work
Bipartite Ranking
Multiview Learning
Semi-supervised Classification
Single View Semi-supervised Ranking
Semi-supervised Multiview Learning for Ranking
Framework
Disagreement for Bipartite Ranking
Algorithm
Weighted Pseudo-labeling
Supervised Ranking Algorithm
Experimental Results
Models
Results
Conclusion
References
Non-redundant Subgroup Discovery in Large and Complex Data
Introduction
Preliminaries
Subgroup Discovery and Exceptional Model Mining
Subgroup Search
Non-Redundant Generalised Subgroup Discovery
Quality Measures
Non-Redundant Beam Selection
Improving Individual Subgroups
Non-Redundant Beam Search
Experiments
A Characteristic Experiment in Detail
Quantitative Results
Related Work
Conclusions
References
Larger Residuals, Less Work: Active Document Scheduling for Latent Dirichlet Allocation
Introduction
Latent Dirichlet Allocation
isLDA Random Matrix Factorization
Residual Latent Dirichlet Allocation
Experimental Evaluation
Conclusions
References
A Community-Based Pseudolikelihood Approach for Relationship Labeling in Social Networks
Introduction
Related Work
Approach
Step 1: Community Detection
Step 2: Conditional Random Field Construction
Step 3: The Pseudolikelihood Model
Experiments
Datasets
Experimental Setup
Results and Discussions
Conclusion and Future Work
References
Correcting Bias in Statistical Tests for Network Classifier Evaluation
Introduction
Network Classifier Evaluation
Theoretical Analysis
Analytical Correction for Bias
Experimental Results
Experiments with Simulated Classifiers
Experiments with Real Classifiers
Conclusion
References
Differentiating Code from Data in x86 Binaries
Introduction
A Language Model for Disassembling x86 Executables
Code Segmentation
Context-Based Data Compression Model
Classification
Experimental Results
Tagging Results
Classification Results
eMule Case Study
Conclusion
References
Bayesian Matrix Co-Factorization: Variational Algorithm and Cram´er-Rao Bound
Introduction
Bayesian Matrix Co-Factorization
Updating Factor Matrices
Learning Hyperparameters
Predictive Distribution
BMCF for General Cases
Bayesian Cramér-Rao Bounds for Bayesian Matrix Co-Factorization
Computation of Fisher Information Matrix
Computing Reconstruction Error
Numerical Experiments
BCRB Comparison on Synthetic Data
Collaborative Prediction in the Cold-Start Situation
Conclusions
References
Learning from Inconsistent and Unreliable Annotators by a Gaussian Mixture Model and Bayesian Information Criterion
Introduction
Background
The Gaussian Mixture Model
Bayesian Model Selection
Method
ML Estimation of the Model Parameters
MAP Estimation of the Unknown True Labels
The GMM-MAPML Algorithm
Analysis of the Model
Experimental Results
Emotional Speech Classification Experiment
CASP92 Protein Disorder Prediction Experiment
Conclusion
References
iDVS: An Interactive Multi-document Visual Summarization System
Introduction
Related Work
Multi-document Summarization
Visual Text Analysis
General Visualization Analytic
Semi-Supervised Learning
System Framework
Methodology
Sentence Cluster Layout
Paragraph Selection and Users' Annotation
Semi-Supervised Document Summarization
An Illustrative Example
Experiments
Data Description
Implemented Summarization Systems
Evaluation Method
Experimental Results
A User Survey
Conclusion
References
Discriminative Experimental Design
Introduction
Active Learning and TED
Discriminative Experimental Design
Objective Function
Optimization Procedure
Experiments
Conclusion
References
Active Learning with Evolving Streaming Data
Introduction
Related Work
Strategies
Setting
Random Strategy
Fixed Uncertainty Strategy
Variable Uncertainty Strategy
Uncertainty Strategy with Randomization
Analysis of How the Labeling Strategies Learn
Ability to Learn Changes
Learning in Stationary Situations
Experimental Evaluation
Datasets
Results on Prediction Datasets
Results on Textual Datasets
Efficiency
Conclusion
References
Demo Papers
Celebrity Watch: Browsing News Content by Exploiting Social Intelligence
Introduction
Methods
Discussion and Conclusions
References
MOA: A Real-Time Analytics Open Source Framework
Introduction
Experimental Framework
Website, Tutorials, and Documentation
References
L-SME: A System for Mining Loosely Structured Motifs
Introduction
Motif Discovery Problem
System Functionalities
References
The MASH Project
Introduction
The MASH Platform
Contests
Conclusion
References
Activity Recognition with Mobile Phones
Overview
Background
Implementation
Conclusion
References
MIME: A Framework for Interactive Visual Pattern Mining
Introduction
Description of the System
Related Work
References
InFeRno - An Intelligent Framework for Recognizing Pornographic Web Pages
Introduction
System Architecture
Classification System
References
MetaData Retrieval: A Software Prototype for the Annotation of Maps with Social Metadata
Introduction to MDR and Comparison with Related Works
MDR Description
Architecture Description
Geographic Characterization
References
TRUMIT: A Tool to Support Large-Scale Mining of Text Association Rules
Introduction
Text Association Rule Mining
TRUMIT: Text Rule Mining Testbench
Demonstration
References
Traffic Jams Detection Using Flock Mining
Introduction
Problem Definition
The Case Study
Additional Analyses
References
Exploring City Structure from Georeferenced Photos Using Graph Centrality Measures
Introduction
Visual Analytics Approach
Pre-processing
Aggregation of Trajectories
Graph of Moves and Place Centrality
Analysis of Results
Demonstration: Seattle Metropolitan Area Photos
References
Author Index

Systemvoraussetzungen

Als PDF speichern Als Link merken