Kiri L. Wagstaff

Constrained Clustering: Advances in Algorithms, Theory, and Applications

Since the initial work on constrained clustering, there have been numerous advances in methods, applications, and our understanding of the theoretical properties of constraints and constrained clustering algorithms. Bringing these developments together, Constrained Clustering: Advances in Algorithms, Theory, and Applications presents an extensive collection of the latest innovations in clustering data analysis methods that use background knowledge encoded as constraints. Algorithms The first five chapters of this volume investigate advances in the use of instance-level, pairwise constraints for partitional and hierarchical clustering. The book then explores other types of constraints for clustering, including cluster size balancing, minimum cluster size,and cluster-level relational constraints. Theory It also describes variations of the traditional clustering under constraints problem as well as approximation algorithms with helpful performance guarantees. Applications The book ends by applying clustering with constraints to relational data, privacy-preserving data publishing, and video surveillance data. It discusses an interactive visual clustering approach, a distance metric learning approach, existential constraints, and automatically generated constraints. With contributions from industrial researchers and leading academic experts who pioneered the field, this volume delivers thorough coverage of the capabilities and limitations of constrained clustering methods as well as introduces new types of constraints and clustering algorithms.

Mining GPS Traces for Map Refinement

Despite the increasing popularity of route guidance systems, current digital maps are still inadequate for many advanced applications in automotive safety and convenience. Among the drawbacks are the insufficient accuracy of road geometry and the lack of fine-grained information, such as lane positions and intersection structure. In this paper, we present an approach to induce high-precision maps from traces of vehicles equipped with differential GPS receivers. Since the cost of these systems is rapidly decreasing and wireless technology is advancing to provide the communication infrastructure, we expect that in the next few years large amounts of car data will be available inexpensively. Our approach consists of successive processing steps: individual vehicle trajectories are divided into road segments and intersections; a road centerline is derived for each segment; lane positions are determined by clustering the perpendicular offsets from it; and the transitions of traces between segments are utilized in the generation of intersection models. This paper describes an approach to this complex data-mining task in a contiguous manner. Among the new contributions are a spatial clustering algorithm for inferring the connectivity structure, more powerful lane finding algorithms that are able to handle lane splits and merges, and an approach to inferring detailed intersection models.

Measuring constraint-set utility for partitional clustering algorithms

Clustering with constraints is an active area of machine learning and data mining research. Previous empirical work has convincingly shown that adding constraints to clustering improves performance, with respect to the true data labels. However, in most of these experiments, results are averaged over different randomly chosen constraint sets, thereby masking interesting properties of individual sets. We demonstrate that constraint sets vary significantly in how useful they are for constrained clustering; some constraint sets can actually decrease algorithm performance. We create two quantitative measures, informativeness and coherence, that can be used to identify useful constraint sets. We show that these measures can also help explain differences in performance for four particular constrained clustering algorithms.

Multidocument summarization via information extraction

We present and evaluate the initial version of RIPTIDES, a system that combines information extraction, extraction-based summarization, and natural language generation to support user-directed multidocument summarization.

Alpha seeding for support vector machines

A key practical obstacle in applying support vector machines to many large-scale data mining tasks is that SVMs generally scale quadratically (or worse) in the number of examples or support vectors.

Active constrained clustering by examining spectral eigenvectors

This work focuses on the active selection of pairwise constraints for spectral clustering. We develop and analyze a technique for Active Constrained Clustering by Examining Spectral eigenvectorS (ACCESS) derived from a similarity matrix. The ACCESS method uses an analysis based on the theoretical properties of spectral decomposition to identify data items that are likely to be located on the boundaries of clusters, and for which providing constraints can resolve ambiguity in the cluster descriptions. Empirical results on three synthetic and five real data sets show that ACCESS significantly outperforms constrained spectral clustering using randomly selected constraints.

VAST: An ASKAP survey for variables and slow transients

The Australian Square Kilometre Array Pathfinder (ASKAP) will give us an unprecedented opportunity to investigate the transient sky at radio wavelengths. In this paper we present VAST, an ASKAP survey for Variables and Slow Transients. VAST will exploit the wide-field survey capabilities of ASKAP to enable the discovery and investigation of variable and transient phenomena from the local to the cosmological, including flare stars, intermittent pulsars, X-ray binaries, magnetars, extreme scattering events, interstellar scintillation, radio supernovae, and orphan afterglows of gamma-ray bursts. In addition, it will allow us to probe unexplored regions of parameter space where new classes of transient sources may be detected. In this paper we review the known radio transient and variable populations and the current results from blind radio surveys. We outline a comprehensive program based on a multi-tiered survey strategy to characterise the radio transient sky through detection and monitoring of transient and variable sources on the ASKAP imaging timescales of 5 s and greater. We also present an analysis of the expected source populations that we will be able to detect with VAST.

The Commensal Real-Time ASKAP Fast-Transients (CRAFT) Survey

We are developing a purely commensal survey experiment for fast (<5 s) transient radio sources. Short-timescale transients are associated with the most energetic and brightest single events in the Universe. Our objective is to cover the enormous volume of transients parameter space made available by ASKAP, with an unprecedented combination of sensitivity and field of view. Fast timescale transients open new vistas on the physics of high brightness temperature emission, extreme states of matter and the physics of strong gravitational fields. In addition, the detection of extragalactic objects affords us an entirely new and extremely sensitive probe on the huge reservoir of baryons present in the IGM. We outline here our approach to the considerable challenge involved in detecting fast transients, particularly the development of hardware fast enough to dedisperse and search the ASKAP data stream at or near real-time rates. Through CRAFT, ASKAP will provide the testbed of many of the key technologies and survey modes proposed for high time resolution science with the SKA.

Current results from a rover science data analysis system

The Onboard Autonomous Science Investigation System (OASIS) evaluates geologic data gathered by a planetary rover. This analysis is used to prioritize the data for transmission, so that the data with the highest science value is transmitted to Earth. In addition, the onboard analysis results are used to identify science opportunities. A planning and scheduling component of the system enables the rover to take advantage of the identified science opportunity. OASIS is a NASA-funded research project that is currently being tested on the FIDO rover at JPL for use on future missions. In this paper, we provide a brief overview of the OASIS system, and then describe our recent successes in integrating with and using rover hardware. OASIS currently works in a closed loop fashion with onboard control software (e.g., navigation and vision) and has the ability to autonomously perform the following sequence of steps: analyze gray scale images to find rocks, extract the properties of the rocks, identify rocks of interest, retask the rover to take additional imagery of the identified target and then allow the rover to continue on its original mission. We also describe the early 2004 ground test validation of specific OASIS components on selected Mars exploration rover (MER) images. These components include the rock-finding algorithm, RockIT, and the rock size feature extraction code. Our team also developed the RockIT GUI, an interface that allows users to easily visualize and modify the rock-finder results. This interface has allowed us to conduct preliminary testing and validation of the rock-finders performance.

Value, cost, and sharing: open issues in constrained clustering

Clustering is an important tool for data mining, since it can identify major patterns or trends without any supervision (labeled data). Over the past five years, semi-supervised (constrained) clustering methods have become very popular. These methods began with incorporating pairwise constraints and have developed into more general methods that can learn appropriate distance metrics. However, several important open questions have arisen about which constraints are most useful, how they can be actively acquired, and when and how they should be propagated to neighboring points. This position paper describes these open questions and suggests future directions for constrained clustering research.