| |
| |
| |
| Introduction | |
| |
| |
| |
| A Natural Representation of Data Similarity | |
| |
| |
| |
| Goals | |
| |
| |
| |
| Outline | |
| |
| |
| |
| Basic Structure of High-Dimensional Spaces | |
| |
| |
| |
| Comparing Attributes | |
| |
| |
| |
| Comparing Records | |
| |
| |
| |
| Similarity | |
| |
| |
| |
| High-Dimensional Spaces | |
| |
| |
| |
| Summary | |
| |
| |
| |
| Algorithms | |
| |
| |
| |
| Improving the Natural Geometry | |
| |
| |
| |
| Projection | |
| |
| |
| |
| Singular Value Decompositions | |
| |
| |
| |
| Random Projections | |
| |
| |
| |
| Algorithms that Find Standalone Clusters | |
| |
| |
| |
| Clusters Based on Density | |
| |
| |
| |
| Parallel Coordinates | |
| |
| |
| |
| Independent Component Analysis | |
| |
| |
| |
| Latent Dirichlet Allocation | |
| |
| |
| |
| Algorithms that Find Clusters and Their Relationships | |
| |
| |
| |
| Clusters Based on Distance | |
| |
| |
| |
| Clusters Based on Distribution | |
| |
| |
| |
| Semidiscrete Decomposition | |
| |
| |
| |
| Hierarchical Clustering | |
| |
| |
| |
| Minimum Spanning Tree with Collapsing | |
| |
| |
| |
| Overall Process for Constructing a Skeleton | |
| |
| |
| |
| Algorithms that Wrap Clusters | |
| |
| |
| |
| Distance-Based | |
| |
| |
| |
| Distribution-Based | |
| |
| |
| |
| 1-Class Support Vector Machines | |
| |
| |
| |
| Autoassociative Neural Networks | |
| |
| |
| |
| Covers | |
| |
| |
| |
| Algorithms to Place Boundaries Between Clusters | |
| |
| |
| |
| Support Vector Machines | |
| |
| |
| |
| Random Forests | |
| |
| |
| |
| Overall Process for Constructing Empty Space | |
| |
| |
| |
| Summary | |
| |
| |
| |
| Spaces with a Single Center | |
| |
| |
| |
| Using Distance | |
| |
| |
| |
| Using Density | |
| |
| |
| |
| Understanding the Skeleton | |
| |
| |
| |
| Understanding Empty Space | |
| |
| |
| |
| Summary | |
| |
| |
| |
| Spaces with Multiple Centers | |
| |
| |
| |
| What is a Cluster? | |
| |
| |
| |
| Identifying Clusters | |
| |
| |
| |
| Clusters Known Already | |
| |
| |
| |
| Finding Clusters | |
| |
| |
| |
| Finding the Skeleton | |
| |
| |
| |
| Empty Space | |
| |
| |
| |
| An Outer Boundary and Novel Data | |
| |
| |
| |
| Interesting Data | |
| |
| |
| |
| One-Cluster Boundaries | |
| |
| |
| |
| One-Cluster-Against-the-Rest Boundaries | |
| |
| |
| |
| Summary | |
| |
| |
| |
| Representation by Graphs | |
| |
| |
| |
| Building a Graph from Records | |
| |
| |
| |
| Local Similarities | |
| |
| |
| |
| Embedding Choices | |
| |
| |
| |
| Using the Embedding for Clustering | |
| |
| |
| |
| Summary | |
| |
| |
| |
| Using Models of High-Dimensional Spaces | |
| |
| |
| |
| Understanding Clusters | |
| |
| |
| |
| Structure in the Set of Clusters | |
| |
| |
| |
| Semantic Stratified Sampling | |
| |
| |
| |
| Ranking Using the Skeleton | |
| |
| |
| |
| Ranking Using Empty Space | |
| |
| |
| |
| Applications to Streaming Data | |
| |
| |
| |
| Concealment | |
| |
| |
| |
| Summary | |
| |
| |
| |
| Including Contextual Information | |
| |
| |
| |
| What is Context? | |
| |
| |
| |
| Changing Data | |
| |
| |
| |
| Changing Analyst and Organizational Properties | |
| |
| |
| |
| Changing Algorithmic Properties | |
| |
| |
| |
| Letting Context Change the Models | |
| |
| |
| |
| Recomputing the View | |
| |
| |
| |
| Recomputing Derived Structures | |
| |
| |
| |
| Recomputing the Clustering | |
| |
| |
| |
| Summary | |
| |
| |
| |
| Conclusions | |
| |
| |
| References | |
| |
| |
| Index | |