Search

Boundaries are the key cue to differentiate objects from each other and the background. However whether boundaries can be regarded as such cannot be determined generally as this highly depends on specific questions that need to be answered. As humans are best able to answer these questions and provide the required knowledge, it is often necessary to learn task-specific boundary properties from user-provided examples. However, current approaches to learning boundaries from examples completely ignore the inherent inaccuracy of human boundary tracings and, hence, derive an imprecise boundary description. We therefore provide an alternative view on supervised boundary learning and propose an efficient and robust algorithm to derive a precise boundary model for boundary detection.

In this paper we consider classification tasks where the class depends on the co-evolution of multiple variables over time, for instance, “if A happens before B and in the meantime we do not observe C, then we have a failure of class X”. We present a two-phased approach to derive such patterns from data. In the first step, we seek the most specific pattern that still matches all data from one class and in the second step we constrain the pattern further, such that it discriminates with respect to other classes. While the second step is directly motivated by the classification task, the first step enables the user to better match his or her mental model of the temporal process to the patterns derived by the classifier. The experimental evaluation on the libras dataset has shown that the additional first step not only improves the interpretability, but also the classification results.

We present OpenML, a novel open science platform that provides easy access to machine learning data, software and results to encourage further study and application. It organizes all submitted results online so they can be easily found and reused, and features a web API which is being integrated in popular machine learning tools such as Weka, KNIME, RapidMiner and R packages, so that experiments can be shared easily.

The integration of heterogeneous data from various domains without the need for prefiltering prepares the ground for bisociative knowledge discoveries where attempts are made to find unexpected relations across seemingly unrelated domains. Information networks, due to their flexible data structure, lend themselves perfectly to the integration of these heterogeneous data sources. This chapter provides an overview of different types of information networks and categorizes them by identifying several key properties of information units and relations which reflect the expressiveness and thus ability of an information network to model heterogeneous data from diverse domains. The chapter progresses by describing a new type of information network known as bisociative information networks. This kind of network combines the key properties of existing networks in order to provide the foundation for bisociative knowledge discoveries. Finally based on this data structure three different patterns are described that fulfill the requirements of a bisociation by connecting concepts from seemingly unrelated domains.

Modern knowledge discovery methods enable users to discover complex patterns of various types in large information repositories. However, the underlying assumption has always been that the data to which the methods are applied to originates from one domain. The focus of this book, and the BISON project from which the contributions are originating, is a network based integration of various types of data repositories and the development of new ways to analyse and explore the resulting gigantic information networks. Instead of finding well defined global or local patterns they wanted to find domain bridging associations which are, by definition, not well defined since they will be especially interesting if they are sparse and have not been encountered before. The 32 contributions presented in this state-of-the-art volume together with a detailed introduction to the book are organized in topical sections on bisociation; representation and network creation; network analysis; exploration; and applications and evaluation.

Data mining started off by finding clearly defined patterns in large sets of relatively homogeneous data. Over the years, increasingly complex data sources were tackled. As a result, newly developed methods grew in complexity, but the basic assumption that the type of pattern sought for was known beforehand remained a constant. I argue that we will ultimately require new systems which enable users to gain new, often surprising insights before they can even determine how to fine-tune and/or validate the patterns themselves.

Knowledge discovery generally focuses on finding patterns within a reasonably well connected domain of interest. In this article we outline a framework for the discovery of new connections between domains (so called bisociations), supporting the creative discovery process in a more powerful way. We motivate this approach, show the difference to classical data analysis and conclude by describing a number of different types of domain-crossing connections.

The discovery of surprising relations in large, heterogeneous information repositories is gaining increasing importance in real world data analysis. If these repositories come from diverse origins, forming different domains, domain bridging associations between otherwise weakly connected domains can provide insights into the data that can otherwise not be accomplished. In this paper, we propose a first formalization for the detection of such potentially interesting, domain-crossing relations based purely on structural properties of a relational knowledge description.

The IDA Symposium series had evolved into a meeting apparently concerned with routine data mining. This report describes our attempts to re-focus the ninth IDA symposium on objectives derived from the genesis of the series. These attempts included a dramatically modified call for papers, and a drastically changed reviewing procedure.

Changing the Focus of the IDA Symposium.- Changing the Focus of the IDA Symposium.- Invited Papers.- Graph Identification.- Intelligent Data Analysis of Intelligent Systems.- Selected Contributions.- Measurement and Dynamical Analysis of Computer Performance Data.- Recursive Sequence Mining to Discover Named Entity Relations.- Integration and Dissemination of Citizen Reported and Seismically Derived Earthquake Information via Social Network Technologies.- Detecting Leukaemia (AML) Blood Cells Using Cellular Automata and Heuristic Search.- Oracle Coached Decision Trees and Lists.- Statistical Modelling for Data from Experiments with Short Hairpin RNAs.- InfraWatch: Data Management of Large Systems for Monitoring Infrastructural Performance.- Deterministic Finite Automata in the Detection of EEG Spikes and Seizures.- Bipartite Graphs for Monitoring Clusters Transitions.- Data Mining for Modeling Chiller Systems in Data Centers.- The Applications of Artificial Neural Networks in the Identification of Quantitative Structure-Activity Relationships for Chemotherapeutic Drug Carcinogenicity.- Image Approach towards Document Mining in Neuroscientific Publications.- Similarity Kernels for Nearest Neighbor-Based Outlier Detection.- End-to-End Support for Dating Paleolandforms.- Spatial Variable Importance Assessment for Yield Prediction in Precision Agriculture.- Selecting the Links in BisoNets Generated from Document Collections.- Novelty Detection in Projected Spaces for Structural Health Monitoring.- A Framework for Path-Oriented Network Simplification.- A Data-Driven Paradigm to Understand Multimodal Communication in Human-Human and Human-Robot Interaction.- Using CAPTCHAs to Index Cultural Artifacts.

Invited Papers.- On Iterative Algorithms with an Information Geometry Background.- Visual Analytics: Combining Automated Discovery with Interactive Visualizations.- Some Mathematics Behind Graph Property Testing.- Finding Total and Partial Orders from Data for Seriation.- Computational Models of Neural Representations in the Human Brain.- Learning.- Unsupervised Classifier Selection Based on Two-Sample Test.- An Empirical Investigation of the Trade-Off between Consistency and Coverage in Rule Learning Heuristics.- Learning Model Trees from Data Streams.- Empirical Asymmetric Selective Transfer in Multi-objective Decision Trees.- Ensemble-Trees: Leveraging Ensemble Power Inside Decision Trees.- A Comparison between Neural Network Methods for Learning Aggregate Functions.- Feature Selection.- Smoothed Prediction of the Onset of Tree Stem Radius Increase Based on Temperature Patterns.- Feature Selection in Taxonomies with Applications to Paleontology.- Associations.- Deduction Schemes for Association Rules.- Constructing Iceberg Lattices from Frequent Closures Using Generators.- Discovery Processes.- Learning from Each Other.- Comparative Evaluation of Two Systems for the Visual Navigation of Encyclopedia Knowledge Spaces.- A Framework for Knowledge Discovery in a Society of Agents.- Learning and Chemistry.- Active Learning for High Throughput Screening.- An Efficiently Computable Graph-Based Metric for the Classification of Small Molecules.- Mining Intervals of Graphs to Extract Characteristic Reaction Patterns.- Clustering.- Refining Pairwise Similarity Matrix for Cluster Ensemble Problem with Cluster Relations.- Input Noise Robustness and Sensitivity Analysis to Improve Large Datasets Clustering by Using the GRID.- An Integrated Graph and Probability Based Clustering Framework for Sequential Data.- Cluster Analysis in Remote Sensing Spectral Imagery through Graph Representation and Advanced SOM Visualization.- Structured Data.- Mining Unordered Distance-Constrained Embedded Subtrees.- Finding Frequent Patterns from Compressed Tree-Structured Data.- A Modeling Approach Using Multiple Graphs for Semi-Supervised Learning.- Text Analysis.- String Kernels Based on Variable-Length-Don't-Care Patterns.- Unsupervised Spam Detection by Document Complexity Estimation.- A Probabilistic Neighbourhood Translation Approach for Non-standard Text Categorisation.

Genomics.- Improved Robustness in Time Series Analysis of Gene Expression Data by Polynomial Model Based Clustering.- A Hybrid Grid and Its Application to Orthologous Groups Clustering.- Promoter Prediction Using Physico-Chemical Properties of DNA.- Parametric Spectral Analysis of Malaria Gene Expression Time Series Data.- An Efficient Algorithm for Finding Long Conserved Regions Between Genes.- The Reversal Median Problem, Common Intervals, and Mitochondrial Gene Orders.- Data Mining.- Building Structure-Property Predictive Models Using Data Assimilation.- Set-Oriented Dimension Reduction: Localizing Principal Component Analysis Via Hidden Markov Models.- Relational Subgroup Discovery for Descriptive Analysis of Microarray Data.- Applicability of Loop Recombination in Ciliates Using the Breakpoint Graph.- High-Throughput Identification of Chemistry in Life Science Texts.- Beating the Noise: New Statistical Methods for Detecting Signals in MALDI-TOF Spectra Below Noise Level.- Molecular Simulation.- Dynamic Complexity of Chaotic Transitions in High-Dimensional Classical Dynamics: Leu-Enkephalin Folding.- Solvent Effects and Conformational Stability of a Tripeptide.- Grid Assisted Ensemble Molecular Dynamics Simulations of HIV-1 Proteases Reveal Novel Conformations of the Inhibitor Saquinavir.- Molecular Informatics.- A Structure-Based Analysis of Single Molecule Force Spectroscopy (SMFS) Data for Bacteriorhodopsin and Four Mutants.- Classifying the World Anti-Doping Agency's 2005 Prohibited List Using the Chemistry Development Kit Fingerprint.- A Point-Matching Based Algorithm for 3D Surface Alignment of Drug-Sized Molecules.- Systems Biology.- Adaptive Approach for Modelling Variability in Pharmacokinetics.- A New Approach to Flux Coupling Analysis of Metabolic Networks.- Biological Networks / Metabolism.- Software Supported Modelling in Pharmacokinetics.- On the Interpretation of High Throughput MS Based Metabolomics Fingerprints with Random Forest.- Construction of Correlation Networks with Explicit Time-Slices Using Time-Lagged, Variable Interval Standard and Partial Correlation Coefficients.- Computational Neuroscience.- The Language of Cortical Dynamics.- A Simple Method to Simultaneously Track the Numbers of Expressed Channel Proteins in a Neuron.

Data from various areas of Life Sciences have increasingly caught the attention of data mining and machine learning researchers. Not only is the amount of data available mind-boggling but the diverse and heterogenous nature of the information is far beyond any other data analysis problem so far. In sharp contrast to classical data analysis scenarios, the life science area poses challenges of a rather different nature for mainly two reasons. Firstly, the available data stems from heterogenous information sources of varying degrees of reliability and quality and is, without the interactive, constant interpretation of a domain expert, not useful. Furthermore, predictive models are of only marginal interest to those users – instead they hope for new insights into a complex, biological system that is only partially represented within that data anyway. In this scenario, the data serves mainly to create new insights and generate new ideas that can be tested. Secondly, the notion of feature space and the accompanying measures of similarity cannot be taken for granted. Similarity measures become context dependent and it is often the case that within one analysis task several different ways of describing the objects of interest or measuring similarity between them matter. Some more recently published work in the data analysis area has started to address some of these issues. For example, data analysis in parallel universes[1], that is, the detection of patterns of interest in various different descriptor spaces at the same time, and mining of frequent, discriminative fragments in large, molecular data bases[2]. In both cases, sheer numerical performance is not the focus; it is rather the discovery of interpretable pieces of evidence that lights up new ideas in the users mind. Future work in data analysis in the life sciences needs to keep this in mind: the goal is to trigger new ideas and stimulate interesting associations.

Structured data represented in the form of graphs arises in\ud several fields of the science and the growing amount of available data makes distributed graph mining techniques particularly relevant. In this paper, we present a distributed approach to the frequent subgraph mining\ud problem to discover interesting patterns in molecular compounds. The problem is characterized by a highly irregular search tree, whereby no reliable workload prediction is available. We describe the three main\ud aspects of the proposed distributed algorithm, namely a dynamic partitioning of the search space, a distribution process based on a peer-to-peer communication framework, and a novel receiver-initiated, load balancing\ud algorithm. The effectiveness of the distributed method has been evaluated on the well-known National Cancer Institute’s HIV-screening dataset, where the approach attains close-to linear speedup in a network\ud of workstations.

Systems Biology.- Structural Protein Interactions Predict Kinase-Inhibitor Interactions in Upregulated Pancreas Tumour Genes Expression Data.- Biochemical Pathway Analysis via Signature Mining.- Recurrent Neuro-fuzzy Network Models for Reverse Engineering Gene Regulatory Interactions.- Data Analysis and Integration.- Some Applications of Dummy Point Scatterers for Phasing in Macromolecular X-Ray Crystallography.- BioRegistry: A Structured Metadata Repository for Bioinformatic Databases.- Robust Perron Cluster Analysis for Various Applications in Computational Life Science.- Structural Biology.- Multiple Alignment of Protein Structures in Three Dimensions.- Protein Annotation by Secondary Structure Based Alignments (PASSTA).- MAPPIS: Multiple 3D Alignment of Protein-Protein Interfaces.- Genomics.- Frequent Itemsets for Genomic Profiling.- Gene Selection Through Sensitivity Analysis of Support Vector Machines.- The Breakpoint Graph in Ciliates.- Computational Proteomics.- ProSpect: An R Package for Analyzing SELDI Measurements Identifying Protein Biomarkers.- Algorithms for the Automated Absolute Quantification of Diagnostic Markers in Complex Proteomics Samples.- Detection of Protein Assemblies in Crystals.- Molecular Informatics.- Molecular Similarity Searching Using COSMO Screening Charges (COSMO/3PP).- Increasing Diversity in In-silico Screening with Target Flexibility.- Multiple Semi-flexible 3D Superposition of Drug-Sized Molecules.- Molecular Structure Determination and Simulation.- Efficiency Considerations in Solving Smoluchowski Equations for Rough Potentials.- Fast and Accurate Structural RNA Alignment by Progressive Lagrangian Optimization.- Visual Analysis of Molecular Conformations by Means of a Dynamic Density Mixture Model.- Distributed Data Mining.- Distributed BLAST in a Grid Computing Context.- Parallel Tuning of Support Vector Machine Learning Parameters for Large and Unbalanced Data Sets.- The Architecture of a Proteomic Network in the Yeast.

The main task of drug discovery is to find novel bioactive molecules, i.e., chemical compounds that, for example, protect human cells against a virus. One way to support solving this task is to analyze a database of known and tested molecules in order to find structural properties of molecules that determine whether a molecule will be active or inactive, so that future chemical tests can be focused on the most promising candidates. A promising approach to this task was presented in [2]: an algorithm for finding molecular fragments that discriminate between active and inactive molecules. In this paper we review this approach as well as two extensions: a special treatment of rings and a method to find fragments with wildcards based on chemical expert knowledge.

Rule systems have failed to attract much interest in large data analysis problems because they tend to be too simplistic to be use- ful or consist of too many rules for human interpretation. We present a method that constructs a hierarchical rule system, with only a small number of rules at each stage of the hierarchy. Lower levels in this hi- erarchy focus on outliers or areas of the feature space where only weak evidence for a rule was found in the data. Rules further up, at higher lev- els of the hierarchy, describe increasingly general and strongly supported aspects of the data. We demonstrate the proposed method's usefulness on several classification benchmark data sets using a fuzzy rule induction process as the underlying learning algorithm. The results demonstrate how the rule hierarchy allows to build much smaller rule systems and how the model—especially at higher levels of the hierarchy—remains in- terpretable. The presented method can be applied to a variety of local learning systems in a similar fashion.