The monograph "Foundations of Rule Learning" (Springer 2012, 334 pages), co-authored by Prof. Nada Lavrač, lead of the research group Knowledge Technologies is a result of a long research work in the area of machine learning. This book presents the fundamentals of rule learning as investigated in classical machine learning and modern data mining. It can also be used as a textbook for teaching machine learning, as well as a comprehensive referenc to reseach in the field of inductive rule learning. Parts of the book are available on Springer website http://link.springer.com/book/10.1007/978-3-540-75197-7/
COBISS.SI-ID: 26327591
Plant defence signalling response against various pathogens, including viruses, is a complex phenomenon. This work focuses on signalling pathways crucial for the plant defence response, i.e., the salicylic acid, jasmonic acid and ethylene signal transduction pathways, in the Arabidopsis thaliana model plant. The initial signalling network topology was constructed manually by defining the representation formalism, encoding the information from public databases and literature, and composing a pathway diagram. The manually constructed network structure consists of 175 components and 387 reactions. In order to complement the network topology with possibly missing relations, a new approach to automated information extraction from biological literature was developed. This approach, named Bio3graph, allows for automated extraction of biological relations from the literature, resulting in a set of (component1, reaction, component2) triplets and composing a graph structure which can be visualised, compared to the manually constructed topology and examined by the experts. Using a plant defence response vocabulary of components and reaction types, Bio3graph was applied to a set of 9,586 relevant full text articles, resulting in 137 newly detected reactions between the components. Finally, the manually constructed topology and the new reactions were merged to form a network structure consisting of 175 components and 524 reactions. The resulting pathway diagram of plant defence signalling represents a valuable source for further computational modelling and interpretation of omics data.
COBISS.SI-ID: 26363431
We developed a method that properly deals with autocorrelation in data that are not independently and identically distributed (i.i.d.) and provides a multi-level insight into the autocorrelation phenomenon. The method is based on the concept of predictive clustering trees (PCTs) and works for different predictive modeling tasks, including classification and regression, as well as some clustering tasks. We applied this method to several real world problems of spatial regression and classification, as well as problems of network regression coming from the areas of social and spatial networks. The resulting models adapt to local properties of the data, providing, at the same time, smoothed predictions.
COBISS.SI-ID: 26073895
High-dimensional data are by their very nature often difficult to handle by conventional machine-learning algorithms, which is usually characterized as an aspect of the curse of dimensionality. However, it was shown that some of the arising high-dimensional phenomena can be exploited to increase algorithm accuracy. One such phenomenon is hubness, which refers to the emergence of hubs in high-dimensional spaces, where hubs are influential points included in many k-neighbor sets of other points in the data. This phenomenon was previously used to devise a crisp weighted voting scheme for the k-nearest neighbor classifier. In this paper we go a step further by embracing the soft approach, and propose several fuzzy measures for k-nearest neighbor classification, all based on hubness, which express fuzziness of elements appearing in k-neighborhoods of other points. Experimental evaluation on real data from the UCI repository and the image domain suggests that the fuzzy approach provides a useful measure of confidence in the predicted labels, resulting in improvement over the crisp weighted method, as well the standard k-NN classifier.
COBISS.SI-ID: 26382887
The open access book is the result of successful EU FP7 Project BISON. Members of the research program contributed seven book chapters addressing methods for discovering new, domain bridging connections and patterns from heterogeneous data sources. The book is available on Springer website http://www.springer.com/computer/ai/book/978-3-642-31829-0.
COBISS.SI-ID: 25939751