Ioannis P. Vlahavas

Mining Multi-label Data

A large body of research in supervised learning deals with the analysis of single-label data, where training examples are associated with a single label λ from a set of disjoint labels L. However, training examples in several application domains are often associated with a set of labels Y ⊆ L. Such data are called multi-label.

Random k-Labelsets: An Ensemble Method for Multilabel Classification

This paper proposes an ensemble method for multilabel classification. The RAndom k-labELsets (RAKEL) algorithm constructs each member of the ensemble by considering a small random subset of labels and learning a single-label classifier for the prediction of each element in the powerset of this subset. In this way, the proposed algorithm aims to take into account label correlations using single-label classifiers that are applied on subtasks with manageable number of labels and adequate number of examples per label. Experimental results on common multilabel domains involving protein, document and scene classification show that better performance can be achieved compared to popular multilabel classification approaches.

Random k-Labelsets for Multilabel Classification

A simple yet effective multilabel learning method, called label powerset (LP), considers each distinct combination of labels that exist in the training set as a different class value of a single-label classification task. The computational efficiency and predictive performance of LP is challenged by application domains with large number of labels and training examples. In these cases, the number of classes may become very large and at the same time many classes are associated with very few training examples. To deal with these problems, this paper proposes breaking the initial set of labels into a number of small random subsets, called labelsets and employing LP to train a corresponding classifier. The labelsets can be either disjoint or overlapping depending on which of two strategies is used to construct them. The proposed method is called RAkEL (RAndom k labELsets), where k is a parameter that specifies the size of the subsets. Empirical evidence indicates that RAkEL manages to improve substantially over LP, especially in domains with large number of labels and exhibits competitive performance against other high-performing multilabel learning methods.

Protein classification with multiple algorithms

Nowadays, the number of protein sequences being stored in central protein databases from labs all over the world is constantly increasing. From these proteins only a fraction has been experimentally analyzed in order to detect their structure and hence their function in the corresponding organism. The reason is that experimental determination of structure is labor-intensive and quite time-consuming. Therefore there is the need for automated tools that can classify new proteins to structural families. This paper presents a comparative evaluation of several algorithms that learn such classification models from data concerning patterns of proteins with known structure. In addition, several approaches that combine multiple learning algorithms to increase the accuracy of predictions are evaluated. The results of the experiments provide insights that can help biologists and computer scientists design high-performance protein classification systems of high quality.

Tracking recurring contexts using ensemble classifiers: an application to email filtering

Concept drift constitutes a challenging problem for the machine learning and data mining community that frequently appears in real world stream classification problems. It is usually defined as the unforeseeable concept change of the target variable in a prediction task. In this paper, we focus on the problem of recurring contexts, a special sub-type of concept drift, that has not yet met the proper attention from the research community. In the case of recurring contexts, concepts may re-appear in future and thus older classification models might be beneficial for future classifications. We propose a general framework for classifying data streams by exploiting stream clustering in order to dynamically build and update an ensemble of incremental classifiers. To achieve this, a transformation function that maps batches of examples into a new conceptual representation model is proposed. The clustering algorithm is then applied in order to group batches of examples into concepts and identify recurring contexts. The ensemble is produced by creating and maintaining an incremental classifier for every concept discovered in the data stream. An experimental study is performed using (a) two new real-world concept drifting datasets from the email domain, (b) an instantiation of the proposed framework and (c) five methods for dealing with drifting concepts. Results indicate the effectiveness of the proposed representation and the suitability of the concept-specific classifiers for problems with recurring contexts.

An Empirical Study of Lazy Multilabel Classification Algorithms

Multilabel classification is a rapidly developing field of machine learning. Despite its short life, various methods for solving the task of multilabel classification have been proposed. In this paper we focus on a subset of these methods that adopt a lazy learning approach and are based on the traditional k-nearest neighbor (k NN) algorithm. Two are our main contributions. Firstly, we implement BRk NN, an adaptation of the k NN algorithm for multilabel classification that is conceptually equivalent to using the popular Binary Relevance problem transformation method in conjunction with the k NN algorithm, but much faster. We also identify two useful extensions of BRk NN that improve its overall predictive performance. Secondly, we compare this method against two other lazy multilabel classification methods, in order to determine the overall best performer. Experiments on different real-world multilabel datasets, using a variety of evaluation metrics, expose the advantages and limitations of each method with respect to specific dataset characteristics.

A Defeasible Logic Reasoner for the Semantic Web

Defeasible reasoning is a rule-based approach for efficient reasoning with incomplete and inconsistent information. Such reasoning is, among others, useful for ontology integration, where conflicting information arises naturally; and for the modeling of business rules and policies, where rules with exceptions are often used. This paper describes these scenarios in more detail, and reports on the implementation of a system for defeasible reasoning on the Web. The system is called DR-DEVICE and is capable of reasoning about RDF metadata over multiple Web sources using defeasible logic rules. The system is implemented on top of CLIPS production rule system and builds upon R-DEVICE, an earlier deductive rule system over RDF metadata that also supports derived attribute and aggregate attribute rules. Rules can be expressed either in a native CLIPS-like language, or in an extension of the OO-RuleML syntax. The operational semantics of defeasible logic are implemented through compilation into the generic rule language of R-DEVICE. The paper also briefly presents a semantic web broker example for apartment renting.

A Comprehensive Study Over VLAD and Product Quantization in Large-Scale Image Retrieval

This paper deals with content-based large-scale image retrieval using the state-of-the-art framework of VLAD and Product Quantization proposed by Jegou as a starting point. Demonstrating an excellent accuracy-efficiency trade-off, this framework has attracted increased attention from the community and numerous extensions have been proposed. In this work, we make an in-depth analysis of the framework that aims at increasing our understanding of its different processing steps and boosting its overall performance. Our analysis involves the evaluation of numerous extensions (both existing and novel) as well as the study of the effects of several unexplored parameters. We specifically focus on: a) employing more efficient and discriminative local features; b) improving the quality of the aggregated representation; and c) optimizing the indexing scheme. Our thorough experimental evaluation provides new insights into extensions that consistently contribute, and others that do not, to performance improvement, and sheds light onto the effects of previously unexplored parameters of the framework. As a result, we develop an enhanced framework that significantly outperforms the previous best reported accuracy results on standard benchmarks and is more efficient.

An Integrated Approach to Automated Semantic Web Service Composition through Planning

The paper presents an integrated approach for automated semantic web service composition using AI planning techniques. An important advantage of this approach is that the composition process, as well as the discovery of the atomic services that take part in the composition, are significantly facilitated by the incorporation of semantic information. OWL-S web service descriptions are transformed into a planning problem described in a standardized fashion using PDDL, while semantic information is used for the enhancement of the composition process as well as for approximating the optimal composite service when exact solutions are not found. Solving, visualization, manipulation, and evaluation of the produced composite services are accomplished, while, unlike other systems, independence from specific planners is maintained. Implementation was performed through the development and integration of two software systems, namely PORSCE II and VLEPPO. PORSCE II is responsible for the transformation process, semantic enhancement, and management of the results. VLEPPO is a general-purpose planning system used to automatically acquire solutions for the problem by invoking external planners. A case study is also presented to demonstrate the functionality, performance, and potential of the approach.

An Ensemble Pruning Primer

Ensemble pruning deals with the reduction of an ensemble of predictive models in order to improve its efficiency and predictive performance. The last 12 years a large number of ensemble pruning methods have been proposed. This work proposes a taxonomy for their organization and reviews important representative methods of each category. It abstracts their key components and discusses their main advantages and disadvantages. We hope that this work will serve as a good starting point and reference for researchers working on the development of new ensemble pruning methods.