eHumanities Intelligent Analysis and Information System for Humanities and Culture.

On%2DLine Handwriting Recognition with Parallelized Machine Learning Algorithms.

Frequent subgraph mining in outerplanar graphs.
In recent years there has been an increased interest in frequent pattern discovery in large databases of graph structured objects. While the frequent connected subgraph mining problem for tree datasets can be solved in incremental polynomial time, it becomes intractable for arbitrary graph databases. Existing approaches have therefore resorted to various heuristic strategies and restrictions of the search space, but have not identified a practically relevant tractable graph class beyond trees. In this paper, we consider the class of outerplanar graphs, a strict generalization of trees, develop a frequent subgraph mining algorithm for outerplanar graphs, and show that it works in incremental polynomial time for the practically relevant subclass of well-behaved outerplanar graphs, i.e., which have only polynomially many simple cycles. We evaluate the algorithm empirically on chemo- and bioinformatics applications. 1
Listing closed sets of strongly accessible set systems with applications to data mining.

Movement Data Anonymity through Generalization.

Toolkit%2DBased High%2DPerformance Data Mining of Large Data on MapReduce Clusters.



Context%2DBased Clustering of Image Search Results.

Efficient Discovery of Interesting Patterns Based on Strong Closedness.

Efficient discovery of interesting patterns based on strong closedness.

Tight Optimistic Estimates for Fast Subgroup Discovery.
Abstract. Subgroup discovery is the task of finding subgroups of a population which exhibit both distributional unusualness and high generality. Due to the non monotonicity of the corresponding evaluation functions, standard pruning techniques cannot be used for subgroup discovery, requiring the use of optimistic estimate techniques instead. So far, however, optimistic estimate pruning has only been considered for the extremely simple case of a binary target attribute and up to now no attempt was made to move beyond suboptimal heuristic optimistic estimates. In this paper, we show that optimistic estimate pruning can be developed into a sound and highly effective pruning approach for subgroup discovery. Based on a precise definition of optimality we show that previous estimates have been tight only in special cases. Thereafter, we present tight optimistic estimates for the most popular binary and multi-class quality functions, and present a family of increasingly efficient approximations to these optimal functions. As we show in empirical experiments, the use of our newly proposed optimistic estimates can lead to a speed up of an order of magnitude compared to previous approaches. 1
Facilitating Clinico%2DGenomic Knowledge Discovery by Automatic Selection of KDD Processes.
The analysis of clinico-genomic data poses complex problems for machine learning. As high volumes of data can be generated easily, selecting the most suitable KDD-process for the problem at hand becomes increasingly hard, even for experienced researchers. The main idea of this paper is to facilitate process selection by representing each data set by a graph based on the ontology that describes data set attributes, and to apply graph mining methods to perform a similarity search. Some new measures for an effective comparison of a data set graph induced from the ontology are proposed. The effectiveness of the proposed approach is evaluated on three datasets. The results show that using ontology-based characteristics leads to improving the characterization of a data set. 1.
Facilitating Clinico%2DGenomic Knowledge Discovery by Automatic Selection of KDD Processes.
The analysis of clinico-genomic data poses complex problems for machine learning. As high volumes of data can be generated easily, selecting the most suitable KDD-process for the problem at hand becomes increasingly hard, even for experienced researchers. The main idea of this paper is to facilitate process selection by representing each data set by a graph based on the ontology that describes data set attributes, and to apply graph mining methods to perform a similarity search. Some new measures for an effective comparison of a data set graph induced from the ontology are proposed. The effectiveness of the proposed approach is evaluated on three datasets. The results show that using ontology-based characteristics leads to improving the characterization of a data set. 1.
Visual Analytics Methods for Movement Data.
Exploration and analysis of large data sets cannot be carried out using purely visual means but require the involvement of database technologies, computerized data processing, and computational analysis methods. An appropriate combination of these technologies and methods with visualization may facilitate synergetic work of computer and human whereby the unique capabilities of each ‘‘partner’ ’ can be utilized. We suggest a systematic approach to defining what methods and techniques, and what ways of linking them, can appropriately support such a work. The main idea is that software tools prepare and visualize the data so that the human analyst can detect various types of patterns by looking at the visual displays. To facilitate the detection of patterns, we must understand what types of patterns may exist in the data (or, more exactly, in the underlying phenomenon). This study focuses on data describing movements of multiple discrete entities that change their positions in space while preserving their integrity and identity. We define the possible types of patterns in such movement data on the basis of an abstract model of the data as a mathematical function that maps entities and times onto spatial positions. Then, we look for data transformations, computations, and visualization techniques that can facilitate the detection of these types of patterns and are suitable for very large data sets – possibly too large for a computer’s memory. Under such constraints, visualization is applied to data that have previously been aggregated and generalized by means of database operations and/or computational techniques.
Efficient Closed Pattern Mining in Strongly Accessible Set Systems.

Semidefinite Ranking on Graphs.
We consider the problem of ranking the vertices of an undirected graph given some preference relation. Without inconsistent preferences in the data, the preferences would form a partial order and we could aim at finding the linear extension that conforms best with the undirected graph. However, in real data there are also inconsistent preferences and hence we have to allow for a few backward edges. This ‘ranking on graphs ’ problem has been tackled before using spectral relaxations. Recently, it has been shown that semidefinite relaxations offer in many cases better solutions than spectral ones for clustering [1] and transductive classification [2]. In this paper, we investigate semidefinite relaxations of ranking on graphs. We incorporate the preferences by fixing certain angles between the metric embedding of the vertices. The final linear extension is obtained by the random projection method. Experiments on real world data sets show the expected improvements over spectral relaxations. Several special cases of ‘ranking on graphs ’ are well investigated problems like ‘topological sort’, ‘minimum feedback arc set’, and ‘minimum-length ordering’. The latter problems are known to be NP-hard. The best known approximation algorithms for ‘minimum-length ordering ’ are based on semidefinite relaxations and random projection of the resulting point set.
Efficient Closed Pattern Mining in Strongly Accessible Set Systems %28Extended Abstract%29.

Geovisual analytics for spatial decision support Setting the research agenda.

Visual analytics tools for analysis of movement data.

Multi%2Dclass Ensemble%2DBased Active Learning.

Frequent subgraph mining in outerplanar graphs.
In recent years there has been an increased interest in frequent pattern discovery in large databases of graph structured objects. While the frequent connected subgraph mining problem for tree datasets can be solved in incremental polynomial time, it becomes intractable for arbitrary graph databases. Existing approaches have therefore resorted to various heuristic strategies and restrictions of the search space, but have not identified a practically relevant tractable graph class beyond trees. In this paper, we consider the class of outerplanar graphs, a strict generalization of trees, develop a frequent subgraph mining algorithm for outerplanar graphs, and show that it works in incremental polynomial time for the practically relevant subclass of well-behaved outerplanar graphs, i.e., which have only polynomially many simple cycles. We evaluate the algorithm empirically on chemo- and bioinformatics applications. 1
Efficient co%2Dregularised least squares regression.
In many applications, unlabelled examples are inexpensive and easy to obtain. Semisupervised approaches try to utilise such examples to reduce the predictive error. In this paper, we investigate a semi-supervised least squares regression algorithm based on the co-learning approach. Similar to other semisupervised algorithms, our base algorithm has cubic runtime complexity in the number of unlabelled examples. To be able to handle larger sets of unlabelled examples, we devise a semi-parametric variant that scales linearly in the number of unlabelled examples. Experiments show a significant error reduction by co-regularisation and a large runtime improvement for the semi-parametric approximation. Last but not least, we propose a distributed procedure that can be applied without collecting all data at a single site. 1.
Frequent subgraph mining in outerplanar graphs.
In recent years there has been an increased interest in frequent pattern discovery in large databases of graph structured objects. While the frequent connected subgraph mining problem for tree datasets can be solved in incremental polynomial time, it becomes intractable for arbitrary graph databases. Existing approaches have therefore resorted to various heuristic strategies and restrictions of the search space, but have not identified a practically relevant tractable graph class beyond trees. In this paper, we consider the class of outerplanar graphs, a strict generalization of trees, develop a frequent subgraph mining algorithm for outerplanar graphs, and show that it works in incremental polynomial time for the practically relevant subclass of well-behaved outerplanar graphs, i.e., which have only polynomially many simple cycles. We evaluate the algorithm empirically on chemo- and bioinformatics applications. 1


Bias%2DFree Hypothesis Evaluation in Multirelational Domains.

Effective rule induction from labeled graphs.

Bias%2Dfree hypothesis evaluation in multirelational domains.

Machine Learning
The automated categorization (or classification) of texts into predefined categories has witnessed a booming interest in the last ten years, due to the increased availability of documents in digital form and the ensuing need to organize them. In the research community the dominant approach to this problem is based on machine learning techniques: a general inductive process automatically builds a classifier by learning, from a set of preclassified documents, the characteristics of the categories. The advantages of this approach over the knowledge engineering approach (consisting in the manual definition of a classifier by domain experts) are a very good effectiveness, considerable savings in terms of expert labor power, and straightforward portability to different domains. This survey discusses the main approaches to text categorization that fall within the machine learning paradigm. We will discuss in detail issues pertaining to three different problems, namely document representation, classifier construction, and classifier evaluation.
Kernels for Predictive Graph Mining.

Bias%2Dfree Hypothesis Evaluation in Multirelational Domains.

A comparative study on methods for reducing myopia of hill%2Dclimbing search in multirelational learning.
Hill-climbing search is the most commonly  used search algorithm in ILP systems because  it permits the generation of theories in short  running times. However, a well known drawback  of this greedy search strategy is its myopia.
On Graph Kernels Hardness Results and Efficient Alternatives.
Abstract. As most ‘real-world ’ data is structured, research in kernel methods has begun investigating kernels for various kinds of structured data. One of the most widely used tools for modeling structured data are graphs. An interesting and important challenge is thus to investigate kernels on instances that are represented by graphs. So far, only very specific graphs such as trees and strings have been considered. This paper investigates kernels on labeled directed graphs with general structure. It is shown that computing a strictly positive definite graph kernel is at least as hard as solving the graph isomorphism problem. It is also shown that computing an inner product in a feature space indexed by all possible graphs, where each feature counts the number of subgraphs isomorphic to that graph, is NP-hard. On the other hand, inner products in an alternative feature space, based on walks in the graph, can be computed in polynomial time. Such kernels are defined in this paper. 1
Cyclic pattern kernels for predictive graph mining.

On Graph Kernels Hardness Results and Efficient Alternatives.
Abstract. As most ‘real-world ’ data is structured, research in kernel methods has begun investigating kernels for various kinds of structured data. One of the most widely used tools for modeling structured data are graphs. An interesting and important challenge is thus to investigate kernels on instances that are represented by graphs. So far, only very specific graphs such as trees and strings have been considered. This paper investigates kernels on labeled directed graphs with general structure. It is shown that computing a strictly positive definite graph kernel is at least as hard as solving the graph isomorphism problem. It is also shown that computing an inner product in a feature space indexed by all possible graphs, where each feature counts the number of subgraphs isomorphic to that graph, is NP-hard. On the other hand, inner products in an alternative feature space, based on walks in the graph, can be computed in polynomial time. Such kernels are defined in this paper. 1
Learning Minesweeper with Multirelational Learning.
Minesweeper is a one-person game which looks deceptively easy to play, but where average human performance is far from optimal. Playing the game requires logical, arithmetic and probabilistic reasoning based on spatial relationships on the board. Simply checking a board state for consistency is an NP-complete problem. Given the difficulty of hand-crafting strategies to play this and other games, AI researchers have always been interested in automatically learning such strategies from experience. In this paper, we show that when integrating certain techniques into a general purpose learning system (Mio), the resulting system is capable of inducing a Minesweeper playing strategy that beats the winning rate of average human players. In addition, we discuss the necessary background knowledge, present experimental results demonstrating the gain obtained with our techniques and show the strategy learned for the game. 1
A Comparative Evaluation of Feature Set Evolution Strategies for Multirelational Boosting.
Boosting has established itself as a successful  technique for decreasing the generalization  error of classification learners by basing predictions  on ensembles of hypotheses. While  previous research has shown that this technique  can be made to work e#ciently even  in the context of multirelational learning by  using simple learners and active feature selection,  such approaches have relied on simple  and static methods of determining feature  selection ordering a priori and adding  features only in a forward manner. In this  paper, we investigate whether the distributional  information present in boosting can  usefully be exploited in the course of learning  to reweight features and in fact even to dynamically  adapt the feature set by adding the  currently most relevant features and removing  those that are no longer needed. Preliminary  results show that these more informed  feature set evolution strategies surprisingly  have mixed e#ects on the number of features  ultimately used in the ensemble, and on the  resulting classification accuracy.
Comparative Evaluation of Approaches to Propositionalization.
Propositionalization has already been shown to be a promising  approach for robustly and e#ectively handling relational data sets for  knowledge discovery. In this paper, we compare up-to-date methods for  propositionalization from two main groups: logic-oriented and databaseoriented  techniques. Experiments using several learning tasks --- both  ILP benchmarks and tasks from recent international data mining competitions  --- show that both groups have their specific advantages. While  logic-oriented methods can handle complex background knowledge and  provide expressive first-order models, database-oriented methods can be  more e#cient especially on larger data sets. Obtained accuracies vary  such that a combination of the features produced by both groups seems  a further valuable venture.
Multirelational data mining 2003 workshop report.

Multirelational data mining 2003 workshop report.

Feature Selection for Propositionalization.

Scaling Boosting by Margin%2DBased Inclusionof Features and Relations.

Macro%2DOperators in Multirelational Learning A Search%2DSpace Reduction Technique.
Refinement operators are frequently used in the area of multirelational learning (Inductive Logic Programming, ILP) in order to search systematically through a generality order on clauses for a correct theory. Only the clauses reachable by a finite number of applications of a refinement operator are considered by a learning system using this refinement operator; ie. the refinement operator determines the search space of the system. For eciency reasons, we would like a refinement operator to compute the smallest set of clauses necessary to find a correct theory.
On the Stability of Example%2DDriven Learning Systems A Case Study in Multirelational Learning.
A popular way of dealing with the complexity of learning from examples is to proceed in an example-driven fashion. In the past, several researchers have shown that using an example-driven approach, it is possible to learn even structurally complex generalizations which would have been difficult to find using other multirelational learning (ILP) algorithms. On the other hand, it is also well known that the quality of the learning results in example-driven learning may depend on the ordering of the examples; however, such stability issues have received almost no attention. In this paper, we present empirical results in several multirelational application domains to show that instability actually affects the performance of a well-known example-driven ILP system. At the same time, we examine one possible solution to the instability problem, presenting an algorithm which relies on stochastically selected examples and parallel search. We show that our algorithm almost eliminates the instability of example-driven search with limited additional effort.
A Scalable Constant%2DMemory Sampling Algorithm for Pattern Discovery in Large Databases.
Many data mining tasks can be seen as an instance of the problem of finding the most interesting (according to some utility function) patterns in a large database. In recent years, significant progress has been achieved in scaling algorithms for this task to very large databases through the use of sequential sampling techniques. However, except for sampling-based greedy algorithms which cannot give absolute quality guarantees, the scalability of existing approaches to this problem is only with respect to the data, not with respect to the size of the pattern space: it is universally assumed that the entire hypothesis space fits in main memory. In this paper, we describe how this class of algorithms can be extended to hypothesis spaces that do not fit in memory while maintaining the algorithms&#039; precise epsilon - delta quality guarantees. We present a constant memory algorithm for this task and prove that it possesses the required properties. In an empirical comparison, we compare variable memory and constant memory sampling.
Finding the Most Interesting Patterns in a Database Quickly by Using Sequential Sampling.
Many discovery problems, e.g., subgroup or association rule discovery, can naturally be cast as n-best hypotheses problems where the goal is to  nd the n hypotheses from a given hypothesis space that score best according to a certain utility function. We present a sampling algorithm that solves this problem by issuing a small number of database queries while guaranteeing precise bounds on con  dence and quality of solutions. Known sampling approaches have treated single hypothesis selection problems, assuming that the utility be the average (over the examples) of some function | which is not the case for many frequently used utility functions. We show that our algorithm works for all utilities that can be estimated with bounded error. We provide these error bounds and resulting worst-case sample bounds for some of the most frequently used utilities, and prove that there is no sampling algorithm for a popular class of utility functions that cannot be estimated with bounded error. The algorithm is sequential in the sense that it starts to return (or discard) hypotheses that already seem to be particularly good (or bad) after a few examples. Thus, the algorithm is almost always faster than its worst-case bounds.
Lerning Hidden Markov Models for Information Extraction Actively from Partially Labeled Text.

Towards Discovery of Deep and Wide First%2DOrder Structures A Case Study in the Domain of Mutagenicity.

Scalability%2C Search%2C and Sampling From Smart Algorithms to Active Discovery.

Incremental Maximization of Non%2DInstance%2DAveraging Utility Functions with Applications to Knowledge Discovery Problems.
Data-independent sample bounds are known to grossly overestimate the amount of data needed for most individual problem instances. This has led to significant recent interest in sequential algorithms which also give precise guarantees about the quality of results, but determine the amount of data needed based on characteristics of the actual problem instance at hand, and thus need significantly fewer examples. In this paper, we present a practical sequential sampling algorithm which (a) is capable of quickly identifying the n best hypotheses and (b) works for all utility functions that can be estimated with bounded error. The algorithm thus can be used not only for predictive learning, but also for tasks such as association rule finding or subgroup discovery. Our experiments with two real-world domains show that our algorithm can be orders of magnitude faster than non-sequential sampling algorithms.
Mining the Web with Active Hidden Markov Models.
Introduction  Given the enormous amounts of information available only in unstructured or semi-structured textual documents, tools for information extraction (IE) have become enormously important. IE tools identify the relevant information in such documents and convert it into a structured format such as a database or an XML document. While first IE algorithms were hand-crafted sets of rules, researchers soon turned to learning extraction rules from hand-labeled documents. Unfortunately, rule-based approaches sometimes fail to provide the necessary robustness against the inherent variability of document structure, which has led to the recent interest in the use of hidden Markov models (HMMs) [1] for this purpose. Speech recognition and computational biochemistry are well-known applications of HMMs.  Markov model algorithms that are used for part-ofspeech tagging, as well as known hidden Markov models for information extraction [1] require the training documents to be 
Incremental Maximization of Non%2DInstance%2DAveraging Utility Functions with Applications to Knowledge Discovery Problems.
Data-independent sample bounds are known to grossly overestimate the amount of data needed for most individual problem instances. This has led to significant recent interest in sequential algorithms which also give precise guarantees about the quality of results, but determine the amount of data needed based on characteristics of the actual problem instance at hand, and thus need significantly fewer examples. In this paper, we present a practical sequential sampling algorithm which (a) is capable of quickly identifying the n best hypotheses and (b) works for all utility functions that can be estimated with bounded error. The algorithm thus can be used not only for predictive learning, but also for tasks such as association rule finding or subgroup discovery. Our experiments with two real-world domains show that our algorithm can be orders of magnitude faster than non-sequential sampling algorithms.
Active Hidden Markov Models for Information Extraction.

Transformation%2DBased Learning Using Multirelational Aggregation.

Relational Learning Using Constrained Confidence%2DRated Boosting.
In propositional learning, boosting has been a very popular  technique for increasing the accuracy of classification learners. In firstorder  learning, on the other hand, surprisingly little attention has been  paid to boosting, perhaps due to the fact that simple forms of boosting  lead to loss of comprehensibility and are too slow when used with standard  ILP learners. In this paper, we show how both concerns can be addressed  by using a recently proposed technique of constrained confidencerated  boosting and a fast weak ILP learner. We give a detailed description  of our algorithm and show on two standard benchmark problems that  indeed such a weak learner can be boosted to perform comparably to  state-of-the-art ILP systems while maintaining acceptable comprehensibility  and obtaining short run-times.
Scalability%2C Search%2C and Sampling From Smart Algorithms to Active Discovery.

Relational Instance%2DBased Learning with Lists and Terms.

Extending K%2DMeans Clustering to First%2DOrder Representations.

A sequential sampling algorithm for a general class of utility criteria.
Many discovery problems, e.g., subgroup or association rule discovery, can naturally be cast as n-best hypothesis problems where the goal is to nd the n hypotheses from a given hypothesis space that score best according to a given utility function. We present a sampling algorithm that solves this problem by issuing a small number of database queries while guaranteeing precise bounds on condence and quality of solutions. Known sampling algorithms assume that the utility be the average (over the examples) of some function, which is not the case for many frequently used utility functions. We show that our algorithm works for all utilities that can be estimated with bounded error. We provide such error bounds and resulting worst-case sample bounds for some of the most frequently used utilities, and prove that there is no sampling algorithm for another popular class of utility functions. The algorithm is sequential in the sense that it starts to return (or discard) hypotheses that already...
Application of Different Learning Methods to Hungarian Part%2Dof%2DSpeech Tagging.
. From the point of view of computational linguistics, Hungarian  is a difficult language due to its complex grammar and rich morphology.  This means that even a common task such as part-of-speech tagging  presents a new challenge for learning when looked at for the Hungarian  language, especially given the fact that this language has fairly free word  order. In this paper we therefore present a case study designed to illustrate  the potential and limits of current ILP and non-ILP algorithms  on the Hungarian POS-tagging task. We have selected the popular C4.5  and Progol systems as propositional and ILP representatives, adding experiments  with our own methods AGLEARN, a C4.5 preprocessor based  on attribute grammars, and the ILP approaches PHM and RIBL. The  systems were compared on the Hungarian version of the multilingual  morphosyntactically annotated MULTEXT-East TELRI corpus which  consists of about 100.000 tokens. Experimental results indicate that Hungarian  POS-tagging is ind...
Scalability Issues in Inductive Logic Programming.

Measuring similarity of RNA structures by relational instance%2Dbased learning A first step toward detecting RNA signal structures in silico.

Term Comparisons in First%2DOrder Similarity Measures.

Relational Distance%2DBased Clustering.
Work on first-order clustering has primarily been focused on the task of conceptual clustering, i.e., forming clusters with symbolic generalizations in the given representation language. By contrast, for propositional representations, experience has shown that simple algorithms based exclusively on distance measures can often outperform their concept-based counterparts. In this paper, we therefore build on recent advances in the area of #rst-order distance metrics and present RDBC, a bottom-up agglomerative clustering algorithm for #rst-order representations that relies on distance information only and features a novel parameter-free pruning measure for selecting the #nal clustering from the cluster tree. The algorithm can empirically be shown to produce good clusterings #on the mutagenesis domain# that, when used for subsequent prediction tasks, improve on previous clustering results and approach the accuracies of dedicated predictive learners.
Term Comparisons in First%2DOrder Similarity Measures.


Data Mining Serviceteil.


Einsatz von Data Mining%2DTechniken zur Analyse %F6kologischer Standort%2D und Pflanzendaten.

An Algorithm for Multi%2Drelational Discovery of Subgroups.
 We consider the problem of finding statistically unusual subgroups  in a multi-relation database, and extend previous work on singlerelation  subgroup discovery. We give a precise definition of the multirelation  subgroup discovery task, propose a specific form of declarative  bias based on foreign links as a means of specifying the hypothesis space,  and show how propositional evaluation functions can be adapted to the  multi-relation setting. We then describe an algorithm for this problem  setting that uses optimistic estimate and minimal support pruning, an  optimal refinement operator and sampling to ensure efficiency and can  easily be parallelized.  
Interactive Configuration in KIKon.

Data Mining %2D Das aktuelle Schlagwort.

Extensibility in Data Mining Systems.
The successful application of data mining techniques ideally requires both system support for the entire knowledge discovery process and the right analysis algorithms for the particular task at hand. While there are a number of successful data mining systems that support the entire mining process, they usually are limited to a fixed selection of analysis algorithms. In this paper, we argue in favor of extensibility as a key feature of data mining systems, and discuss the requirements that this entails for system architecture. We identify in which points existing data mining systems fail to meet these requirements, and then describe a new integration architecture for data mining systems that addresses these problems based on the concept of &#034;plug-ins&#034;. KEPLER, our data mining system built according to this architecture, is presented and discussed.  Keywords: data mining, system architecture, extensibility,  KEPLER  Introduction  Data Mining, or Knowledge Discovery in Databases (KDD) aims...
Induktive Logikprogrammierung %2D Grundlagen und Techniken.

Machine Learning ECML%2D95
The utility problem in learning systems occurs when knowledge learned in an attempt to improve a system&#039;s performance degrades performance instead. We present a methodology for the analysis of utility problems which uses computational models of problem solving systems to isolate the root causes of a utility problem, to detect the threshold conditions under which the problem will arise, and to design strategies to eliminate it. We present models of case-based reasoning and control-rule learning systems and compare their performance with respect to the swamping utility problem. Our analysis suggests that case-based reasoning systems are more resistant to the utility problem than control-rule learning systems.  1 1. Introduction  An interesting asymmetry exists in the patterns of retrieval in case-based reasoning (CBR) and control-rule learning (CRL) systems: to take advantage of past learning experiences, CRL systems need to retrieve rules from memory at each step, whereas CBR systems ne...
Concept Formation During Interactive Theory Revision.

On the Proper Definition of Minimality in Specialization and Theory Revision.
A central operation in an incremental learning system is the specialization of an incorrect theory in order exclude incorrect inferences. In this paper, we discuss what properties are to be required from such theory revision operations. In particular, we examine what it should mean for a revision to be minimal. As a surprising result, the seemingly most natural criterion, requiring revisions to produce maximally general correct specializations, leads to a number of serious problems. We therefore propose an alternative interpretation of minimality based on the notion of base revision from theory contraction work, and formally define it as a set of base revision postulates. We then present a revision operator (Mbr) that meets these postulates, and shown that it produces the maximally general correct revision satisfying the postulates, i.e., the revisions produced by Mbr are indeed minimal in our sense. The operator is implemented and used in Krt, the knowledge revision tool of the Mobal ...
Panel Evaluating and Changing Representation in Concept Acquisition.

Towards a Model of Grounded Concept Formation.
In most research on concept formation within machine learning and cognitive psychology, the features from which concepts are built are assumed to be provided as elementary vocabulary. In this paper, we argue that this is an unnecessarily limited paradigm within which to examine concept formation. Based on evidence from psychology and machine learning, we contend that a principled account of the origin of features can only be given with a grounded model of concept formation, i.e., with a model that incorporates direct access to the world via sensors and manipulators. We discuss the domain of process control as a suitable framework for research into such models, and present a first approach to the problem of developing elementary vocabularies from perceptual sensor data. 1
Die Umweltverankerung von Begriffsbildungsprozessen.

Automatic Representation Adjustment in an Observational Discovery System.

Design Goals for Sloppy Modeling Systems.

Higher%2Dorder Concepts in a Tractable Knowledge Representation.

Demand%2DDriven Concept Formation.
Most valid rules that are learned from very large and high  dimensional data sets are not interesting, but are already known to the  users. The dominant model of the overall data set may well suppress the  interesting local patterns. The search for interesting local patterns can  be implemented by a two step learning approach which first acquires the  global models before it focuses on the rest in order to detect local patterns.
