Machine Learning and Knowledge Discovery in Databases

Name: Machine Learning and Knowledge Discovery in Databases | European Conference, ECML PKDD 2015, Porto, Portugal, September 7-11, 2015, Proceedings, Part II
Brand: Springer
Price: 53.49 EUR
Availability: OnlineOnly

European Conference, ECML PKDD 2015, Porto, Portugal, September 7-11, 2015, Proceedings, Part II

Annalisa Appice Pedro Pereira Rodrigues Vítor Santos Costa João Gama Alípio Jorge Carlos Soares(Editor)

Springer (Publisher)

Published on 28. August 2015

XLII, 773 pages

E-Book

PDF with digital watermarking

System requirements

978-3-319-23525-7 (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

Intro
Preface
Organization
Abstracts of Journal Track Articles
Contents - Part II
Research Track Matrix and Tensor Analysis
BoostMF: Boosted Matrix Factorisation for Collaborative Ranking
1 Introduction
2 Related Work
3 Boosted Matrix Factorisation (BoostMF)
3.1 Probabilistic Matrix Factorisation (PMF)
3.2 BoostMF
3.3 Theoretical Analysis
4 Experiments
4.1 Datasets and Evaluation Metric
4.2 Experimental Setup
4.3 Results
5 Conclusion
References
Convex Factorization Machines
1 Introduction
2 Factorization Machines
3 Convex Formulation
4 Optimization Algorithm
4.1 Minimizing with Respect to bold0mu mumu ww2005/06/28 ver: 1.3 subfig packagewwww
4.2 Minimizing with Respect to bold0mu mumu ZZ2005/06/28 ver: 1.3 subfig packageZZZZ
4.3 Squared Loss Case
4.4 Computational Complexity
4.5 Convergence Guarantees
5 Experimental Results
5.1 Synthetic Experiments
5.2 Recommender System Experiments
6 Related Work
7 Conclusion
References
Generalized Matrix Factorizations as a Unifying Framework for Pattern Set Mining: Complexity Beyond Blocks
1 Introduction
2 Related Work
3 Definitions
3.1 Generalized Outer Product
3.2 Generalized Rank
4 Computational Complexity
4.1 Rank-1 Submatrices
4.2 Selecting Some Rank-1 Submatrices
4.3 Minimum-Error Sub-Decompositions
4.4 Deciding the Rank
4.5 Minimum-Error Approximate Decompositions
5 Approximability
5.1 Approximating Smallest Sub-Decompositions
5.2 Approximating Minimum-Error Sub-Decompositions
6 Conclusions and Future Work
References
Scalable Bayesian Non-negative Tensor Factorization for Massive Count Data
1 Introduction
2 Canonical PARAFAC Decomposition
3 Beta-Negative Binomial CP Decomposition
3.1 Reparametrizing the Poisson Distribution
4 Inference
4.1 Gibbs Sampling
4.2 Variational Bayes Inference
4.3 Online Inference
Conditional Density Filtering:
Stochastic Variational Inference:
Computational Complexity:
5 Related Work
6 Experiments
6.1 Inferring the Rank
6.2 Tensor Completion Results
6.3 Analyzing Publications Database
6.4 Analyzing Political Science Data
6.5 Analyzing Transactions Data
6.6 Scalability
7 Conclusion
References
A Practical Approach to Reduce the Learning Bias Under Covariate Shift
1 Introduction
2 Preliminaries
3 Problem Analysis
4 Experiments
4.1 Importance Ratio Estimation
4.2 Toy Regression Problem
4.3 Simple Step Sample Selection Distribution
4.4 General Covariate Selection Mechanisms
5 Conclusions
References
Hyperparameter Optimization with Factorized Multilayer Perceptrons
1 Introduction
2 Related Work
3 Background
3.1 Problem Setting
3.2 Requirements for a Surrogate Model
3.3 Proposed Models
3.4 Estimating Prediction Uncertainty
4 Experiments
4.1 Meta Data Set Creation
4.2 Experiment 1: Reconstruction of the Response Surface
4.3 Experiment 2: Uncertainty Estimation in SMBO
4.4 Experiment 3: Sequential Model Based Optimization
5 Conclusions
References
Hyperparameter Search Space Pruning -- A New Component for Sequential Model-Based Hyperparameter Optimization
1 Introduction
1.1 Our Contributions
2 Related Work
3 Background
3.1 The Formal Setup
3.2 Sequential Model-Based Optimization
4 Pruning the Search Space
4.1 Formal Description
5 Experimental Evaluation
5.1 Tuning Strategies
5.2 Evaluation Metrics
5.3 Meta-Data Sets
5.4 Hyperparameter Optimization for SVMs
5.5 Hyperparameter Optimization for Weka
6 Conclusion and Future Work
References
Multi-Task Learning with Group-Specific Feature Space Sharing
1 Introduction
2 Formulation
3 The Proposed Consensus Optimization Algorithm
3.1 Convergence Analysis and Stopping Criteria
3.2 Computational Complexity
4 Generalization Bound Based on Rademacher Complexity
5 Experiments
5.1 Experimental Settings
5.2 Experimental Results
6 Conclusions
References
Opening the Black Box: Revealing Interpretable Sequence Motifs in Kernel-Based Learning Algorithms
1 Introduction
2 Preliminaries
2.1 Weighted-Degree (WD) Kernel
2.2 Positional Oligomer Importance Matrices (POIMs)
2.3 Shortcomings of POIMs
2.4 What is Coming Up: The Proposed Approach in a Nutshell
3 Methodology for Revealing Discriminative Motifs by Mimicking POIMs
3.1 Optimization Problem
3.2 Efficient Computation of motifPOIM
4 Empirical Analysis
4.1 Experimental Setup
4.2 Experimental Results for USPS Dataset
4.3 Results for Synthetic Splice Site Experiments
4.4 Real-World Experiments on Human Splice Data
5 Conclusion and Discussion
References
Pattern and Sequence Mining
Fast Generation of Best Interval Patternsfor Nonmonotonic Constraints
1 Introduction
2 Data Model
2.1 FCA and Pattern Structures
2.2 Interval Pattern Structure
2.3 Stability Index of a Concept
2.4 Projections of Pattern Structures
2.5 Projections of Interval Pattern Structures
3 -o Algorithm
3.1 Anti-monotonicity w.r.t. a Projection
3.2 Anti-monotonicity w.r.t. a Chain of Projections
3.3 Algorithms
3.4 -o Algorithm for Interval Tuple Data
3.5 -o Algorithm for Closed Patterns
3.6 -measure and -o Algorithm
3.7 Example of -Stable Patterns in Interval Tuple Data
4 Experiments and Discussion
4.1 Dataset Simplification
4.2 Datasets
4.3 Experiments
5 Conclusion
References
Non-parametric Jensen-Shannon Divergence
1 Introduction
2 Theory
2.1 Univariate Case
2.2 Multivariate Case
2.3 Computing CJS
2.4 Complexity Analysis
2.5 Summing Up
3 Related Work
4 Experiments
4.1 Statistical Power
4.2 Scalability
4.3 Change Detection on Time Series
4.4 Anomaly Detection on Time Series
4.5 Multivariate Discretisation
4.6 Multi-Target Subgroup Discovery
5 Discussion
6 Conclusion
References
Swap Randomization of Bases of Sequences for Mining Satellite Image Times Series
1 Introduction
2 Grouped Frequent Sequential Patterns
3 Swap Randomization of Base of Sequences Representing SITS
4 GFS-Pattern Assessment and SITS Summarization
5 Experiments
6 Conclusion
References
The Difference and the Norm --- Characterising Similarities and Differences Between Databases
1 Introduction
2 Related Work
3 Preliminaries
3.1 Notation
3.2 MDL, a Brief Primer
4 MDL for the Difference and the Norm
4.1 The Problem, Informally
4.2 Our Models
4.3 Encoded Length of the Data
4.4 Encoded Length of the Model
4.5 The Problem, Formally
5 Algorithm
5.1 The Cover Algorithm
5.2 The DIFFNORM Algorithm
5.3 Candidate Generation and Evaluation
5.4 Estimating Candidate Quality
5.5 Complexity
6 Experiments
6.1 Synthetic Data
6.2 Real World Data
7 Discussion
8 Conclusion
References
Preference Learning and Label Ranking
Dyad Ranking Using A Bilinear Plackett-Luce Model
1 Introduction
2 Dyad Ranking
2.1 Label Ranking
2.2 Dyad Ranking as an Extension of Label Ranking
3 Related Work
4 A Bilinear Plackett-Luce Model
4.1 The Plackett-Luce Model
4.2 Label Ranking Using the PL Model
4.3 Dyad Ranking Using the PL model
4.4 Identifiability of the Bilinear PL Model
4.5 Comparison Between the Linear and Bilinear PL Model
5 Experiments
5.1 Synthetic Data
5.2 Case Study in Meta-Learning
6 Summary and Outlook
References
Fast Training of Support Vector Machines for Survival Analysis
1 Introduction
2 Survival Analysis
3 Survival Analysis as Ranking Problem
3.1 Truncated Newton Optimization
3.2 Efficient Calculation of Search Direction
3.3 Improving Optimization by Order Statistic Trees
4 Survival Analysis as Regression Problem
5 Non-linear Extension
6 Experiments
6.1 Computational Efficiency
6.2 Prediction Performance
7 Conclusion
References
Superset Learning Based on Generalized Loss Minimization
1 Introduction
2 Setting and Notation
3 A Loss Minimization Approach
3.1 Generalized Loss Minimization
3.2 Data Disambiguation
3.3 Examples
3.4 Superset Learning for Structured Output Prediction
4 Label Ranking
4.1 Prediction Accuracy
4.2 Label Ranking Methods
5 Label Ranking based on Labelwise Decomposition
5.1 Complete Training Information
5.2 Incomplete Training Information
5.3 Generalized Nearest Neighbor Estimation
5.4 Experiments
6 Summary and Outlook
Probabilistic, Statistical, and Graphical Approaches
Bayesian Modelling of the Temporal Aspects of Smart Home Activity with Circular Statistics
1 Introduction
2 Related Work
3 Methods
3.1 The Wrapped Normal (WN) Distribution
3.2 Bayesian Inference
3.3 WN Mixture (WNM) Models
3.4 Approximate WN (AWN)
3.5 Approximate WNM (AWNM)
3.6 Model Comparison
3.7 Rose Diagrams
4 Experiments
4.1 Toy Data
4.2 The CASAS HH101 Dataset
4.3 Priors
4.4 Symmetry Breaking
5 Results
5.1 Toy Data
5.2 Smart Home Data
6 Discussion
7 Conclusions
7.1 Further Work
References
Message Scheduling Methods for Belief Propagation
1 Introduction
2 Preliminaries
3 Scheduling
3.1 Residual Belief Propagation
3.2 Noise Injection Belief Propagation
3.3 Weight Decay Belief Propagation
4 Experiments
4.1 Fully Connected Graph with Uniform Parameters
4.2 Ising Grids with Random Factors
4.3 Quality of Marginals
5 Related Work
6 Conclusion
References
Output-Sensitive Adaptive Metropolis-Hastings for Probabilistic Programs
1 Introduction
2 Preliminaries
2.1 Probabilistic Program
2.2 Adaptive Markov Chain Monte Carlo
2.3 Lightweight Metropolis-Hastings
3 Adaptive Lightweight Metropolis-Hastings
3.1 Quantifying Influence
3.2 Propagating Rewards to Variables
4 Convergence of Adaptive LMH
5 Empirical Evaluation
6 Contribution and Future Work
References
Planning in Discrete and Continuous Markov Decision Processes by Probabilistic Programming
1 Introduction
2 Preliminaries
3 Dynamic Distributional Clauses
4 Planning by Importance Sampling
5 Related Work
6 Experiments
7 Practical Improvements
8 Conclusions
References
Simplifying, Regularizing and Strengthening Sum-Product Network Structure Learning
1 Introduction
2 Sum-Product Networks
2.1 Structure Learning
3 Contributions
3.1 Deepening by Limiting Node Splits
3.2 Regularization by Tractable Multivariate Distribution Hybridization
3.3 Strengthening by Model Averaging
4 Related Works
5 Experiments
5.1 Experimental Design
5.2 Results and Discussion
6 Conclusions
References
Sparse Bayesian Recurrent Neural Networks
1 Introduction
2 Recurrent Neural Networks
3 Proposed Approach
3.1 Regression SB-RNN
3.2 Classification SB-RNN
4 Experiments
4.1 Human Motion Modeling
4.2 Acoustic Novelty Detection
4.3 Computational Complexity
5 Conclusions and Future Work
References
Structured Prediction of Sequences and Trees Using Infinite Contexts
1 Introduction
2 Background and Related Work
3 The Model
4 Learning
5 Prediction
5.1 A* Search
5.2 MCMC Sampling
6 Experiments
6.1 Morphological Parsing
6.2 Syntactic Parsing
6.3 Part-of-Speech Tagging
6.4 Analysis
7 Conclusion and Future Work
References
Temporally Coherent Role-Topic Models (TCRTM): Deinterlacing Overlapping Activity Patterns
1 Introduction
2 Related Work
3 Model
3.1 Inference
4 Experiments
4.1 Activity Types
4.2 Perplexity
4.3 Effect of Roles
5 Conclusions
References
The Blind Leading the Blind: Network-Based Location Estimation Under Uncertainty
1 Introduction
2 Related Work
3 Methods
3.1 Problem Setting
3.2 Estimating a Single Location
3.3 Estimating Multiple Dependent Locations
3.4 Baseline Methods
4 Experiments
4.1 Predicting Social Network User Home Locations
4.2 Geotagging Historical Church Records
4.3 Geotagging Flickr Photos
5 Conclusions and Discussion
References
Weighted Rank Correlation: A Flexible Approach Based on Fuzzy Order Relations
1 Introduction
2 Rank Correlation
2.1 Concordance and Discordance
2.2 Rank Correlation Measures
3 Fuzzy Relations
3.1 Fuzzy Equivalence
3.2 Fuzzy Ordering
3.3 Practical Construction
4 Fuzzy Relations on Rank Data
4.1 Scaling Functions on Rank Positions
5 Weighted Rank Correlation
6 Related Work
7 Experiments
7.1 First Study
7.2 Second Study
8 Conclusion and Future Work
References
Rich Data
Concurrent Inference of Topic Models and Distributed Vector Representations
1 Introduction
2 Related Work
3 Problem Formulation
4 Methodology
4.1 Distributed Representation of Heterogeneous Entities
4.2 Estimating Topic Labels of Documents
4.3 Concurrent Training
5 Complexity Analysis
6 Evaluation
7 Experiments
7.1 Analysis of Distributed Representations of Topics and Documents
7.2 Expressiveness of Topic Vectors
7.3 Comparison of Quality of Generated Topics
7.4 Evaluation using Domain Specific Information
7.5 Runtime Characteristics
8 Conclusion
References
Differentially Private Analysis of Outliers
1 Introduction
Related Works.
Our Contribution.
2 Differential Privacy
Global Sensitivity.
Smooth Sensitivity.
3 Problem Statement
3.1 Counting Outliers
3.2 Differential Privacy of Outlier Analysis
4 Differentially Private Count of Outliers
4.1 Difficulties in Global Sensitivity Method
4.2 Local Sensitivity and Smooth Sensitivity
Local Sensitivity.
Smooth Sensitivity.
4.3 Efficient Computation of Smooth Sensitivity Bound
Algorithm for Local Sensitivity Bound.
Algorithm for Smooth Sensiticity Bound.
5 Experiments
5.1 Settings
5.2 Count Outliers
6 Conclusion and Future Works
References
Inferring Unusual Crowd Events from Mobile Phone Call Detail Records
1 Introduction
2 Unusual Event Detection Problem
3 Unusual Event Detection Framework
4 Experiments
4.1 Experimental Setup
4.2 Experimental Results
4.3 Efficiency and Parameters
5 Visual Analytics Prototype System
6 Related Work
7 Conclusion and Future Work
References
Learning Pretopological Spaces for Lexical Taxonomy Acquisition
1 Introduction and Related Work
2 Pretopological Framework
2.1 Pretopology and Multi-criteria Analysis
2.2 Pretopology and LT Acquisition
2.3 Current Limitations
3 Learning Pretopological Spaces
3.1 Parameterized Pretopological Space
3.2 Semi-supervised Learning of P-Spaces
4 Experiments on LT Acquisition
4.1 Experimental Setups
4.2 LT Acquisition with Auto-supervision
4.3 LT Acquisition with Semi-supervision
5 Conclusions
References
Multidimensional Prediction Models When the Resolution Context Changes
1 Introduction
2 Multidimensional Contexts
3 Measure Properties and Mean Models
4 Experimental Setting and Results
5 Related Work
6 Conclusions and Future Work
References
Semi-supervised Subspace Co-Projection for Multi-class Heterogeneous Domain Adaptation
1 Introduction
2 Related Work
3 Semi-supervised Multi-class Heterogeneous Domain Adaptation
3.1 Semi-supervised Learning Framework
3.2 Multi-class Classification with ECOC Schemes
4 Training Algorithm
5 Experiments
5.1 Datasets and Methods
5.2 Cross-lingual Text Classification
5.3 Parameter Sensitivity Analysis
5.4 Experimental Results on UCI Dataset
5.5 Impact of the ECOC Encoding Schemes
6 Conclusion
References
Towards Computation of Novel Ideas from Corpora of Scientific Text
1 Introduction
2 Related Work
3 Defining an ``Idea''
4 Methodology
4.1 Noun-Phrase Extraction
4.2 Noun-Phrase Filtering
4.3 Noun-Phrases Categorization
4.4 Known-Idea Construction
4.5 Relevance Values for Known-Ideas
4.6 Computation of Novel-Idea Pairs
5 Experimental Evaluation
5.1 Results
6 Discussion
7 Conclusion
References
A Appendix
Social and Graphs
Discovering Audience Groups and Group-Specific Influencers
1 Introduction
2 Preliminary
3 The AudClus Method
3.1 Audience Clustering for the Single-direct Case
3.2 Generalized Model
4 Experiment
4.1 Experiment Setup
4.2 Qualitative Analysis and Case Studies
4.3 Quantitative Analysis
4.4 Observations on Group-Specific Influence
5 Related Work
6 Conclusion
References
Estimating Potential Customers Anywhere and Anytime Based on Location-Based Social Networks
1 Introduction
2 Related Work
3 Problem Statement
4 Dataset
5 Potential Customer Estimator (PCE)
5.1 Geographical, Mobility, and Social Features
5.2 Correlation Graph
5.3 Location Correlation
5.4 Customer Inference Algorithm
6 Experiments
6.1 Evaluation Plans
6.2 Experimental Results
7 Conclusion
References
Exact Hybrid Covariance Thresholding for Joint Graphical Lasso
1 Introduction
2 Notation and Definition
3 Joint Graphical Lasso
4 Uniform Thresholding
5 Non-uniform Thresholding
6 Hybrid ADMM (HADMM)
7 Experimental Results
7.1 Correctness of HADMM by Experimental Validation
7.2 Performance on Synthetic Data
7.3 Performance on Real Gene Expression Data
8 Conclusion and Discussion
References
Fast Inbound Top-K Query for Random Walk with Restart
1 Introduction
2 Preliminaries
2.1 Problem Description
2.2 Naïve Methods
2.3 Overview of Squeeze and Ripple
3 The Squeeze Algorithm
4 The Ripple Algorithm
4.1 Algorithm Sketch
4.2 The Lower Bound
4.3 The Upper Bound
5 Experiments
5.1 Experimental Setup
5.2 Illustrating Cases
5.3 Efficiency Study
6 Related Work
7 Conclusions
References
Finding Community Topics and Membership in Graphs
1 Introduction
2 Related Work
3 Background
3.1 Subspace Clustering
3.2 Topic Models
4 Seeded Estimation of Network Communities
4.1 Notation
4.2 Model
4.3 Algorithm
5 Experiments
5.1 Dataset Descriptions
5.2 Methods and Evaluation
5.3 Results
5.4 Interpretation of Detected Communities
6 Conclusion
References
Finding Dense Subgraphs in Relational Graphs
1 Introduction
2 Related Work
2.1 Finding a Dense Subgraph in a Single Graph
2.2 Finding Cross-Graph Quasicliques
3 Methods
3.1 A Greedy Algorithm for Densest Common Subgraph
3.2 Densest Common at Least-k Subgraph (DCalkS)
4 Experiments
4.1 Synthetic Dataset
4.2 Real-World Datasets
5 Discussion
References
Generalized Modularity for Community Detection
1 Introduction
2 Generalized Modularity (GM)
2.1 Comparison to Modularity
2.2 Unified Generalized Modularity (UGM)
2.3 Finding Communities Based on the GM Quality Function
3 Experiments
3.1 Testing Unified Generalized Modularity
3.2 Training Parameters of Generalized Modularity
3.3 Comparison with Other Methods
4 Conclusion
References
Handling Oversampling in Dynamic Networks Using Link Prediction
1 Introduction
1.1 Oversampling
1.2 Link Prediction
1.3 Related Work
2 Problem Formulation and Methodology
3 Generative Models for Graph Sequences
3.1 Generative Model for the Ground-Truth Graph Sequence
3.2 Generative Model for Oversampling
4 Results for Synthetic Data
5 Results for Real-World Data
5.1 Haggle Infocom
5.2 MIT Reality Mining
6 Conclusions
References
Hierarchical Sparse Dictionary Learning
1 Introduction
2 Hierarchical Sparse Structures on Dictionaries
3 Hierarchical Sparse Dictionary Learning
3.1 Approximated sparsity of D on
3.2 Optimization Algorithm
3.3 Analysis of HiSDL Algorithm
4 Related Work
5 Experimental Results
5.1 Evaluation on Empirical Errors
5.2 Evaluation on Atom Recovery
5.3 Evaluation on Sparse Codes
6 Conclusion
References
Latent Factors Meet Homophily in Diffusion Modelling
1 Introduction
2 Related Works
2.1 Latent Factor Models
2.2 Social Influence and Diffusion Models
3 Proposed Framework and Model
3.1 Basic Notations
3.2 Framework
3.3 Topic Interaction and Homophily Aware Diffusion (TIHAD) Model
3.4 Linear Threshold with Latent Factors (LTLF)
4 Learning of TIHAD Model
4.1 Optimization Formulation
4.2 Optimization Solution
5 Experiments
5.1 Impact of Homophily on Diffusion
5.2 Impact of Item Interaction on Diffusion
5.3 Hashtag Diffusion Prediction Evaluation
6 Conclusion
References
Maintaining Sliding-Window Neighborhood Profiles in Interaction Networks
1 Introduction
2 Preliminaries
3 Problem Statement
4 Maintaining the Exact Neighborhood Profile
4.1 Summary for Neighborhood Functions
4.2 Updating Summaries
5 Approximating Neighborhood Function
5.1 Hyperloglog and Sliding-Window Hyperloglog Sketches
5.2 Computation of Neighborhood Profiles Based on Sliding HLL
6 Related Work
7 Experimental Evaluation
8 Concluding Remarks
References
Response-Guided Community Detection: Application to Climate Index Discovery
1 Introduction
2 Response-Guided Community Detection
2.1 Problem Statement
2.2 Algorithms for Response-Guided Community Detection
3 Climate Index Discovery
3.1 Network Construction Methodology
4 Experimental Evaluation
4.1 Data Description
4.2 Data Preprocessing
4.3 Climate Networks Constructed
4.4 Climate Indices Discovered
4.5 Seasonal Rainfall Prediction
4.6 Physical Interpretation of Climate Indices Discovered
5 Conclusions
References
Robust Classification of Information Networks by Consistent Graph Learning
1 Introduction
2 The Proposed Method
2.1 Preliminary Definitions
2.2 Notation
2.3 Standard Graph Regularization
2.4 Consistent Graph Learning
2.5 Optimization
2.6 Estimation of Unlabeled Data
2.7 Analysis
3 Related Work
4 Experiments
4.1 Data Sets
4.2 Baselines and Parameter Settings
4.3 Classification Results on Cora
4.4 Classification Results on DBLP
5 Conclusions and Future Work
References
Author Index

System requirements

Save as PDF Copy link into clipboard

Schweitzer Fachinformationen

Machine Learning and Knowledge Discovery in Databases

Description

More details

Other editions

Additional editions

Content

System requirements