Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

681.3*I26 <043> --- Academic collection --- 681.3*I26 <043> Learning: analogies; concept learning; induction; knowledge acquisition; language acquisition; parameter learning (Artificial intelligence)--See also {681.3*K32}--Dissertaties --- Learning: analogies; concept learning; induction; knowledge acquisition; language acquisition; parameter learning (Artificial intelligence)--See also {681.3*K32}--Dissertaties --- Theses

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Choose an application

• Reference Manager
• EndNote
• RefWorks (Direct export to RefWorks)

Statistical relational learning (SRL) combines state-of-the-art statistical modeling with relational representations. It thereby promises to provide effective machine learning techniques for domains that cannot adequately be described using a propositional representation. Driven by new applications in which data is structured, interrelated, and heterogeneous, this area of machine learning has recently received increasing attention. However, combining statistical modeling and relational representations also poses new challenges. There is a trade-off between the expressivity of a machine learning formalism and its computational efficiency, as a higher expressivity entails a larger search space during learning. Propositional machine learning techniques are at one end of this trade-off, while approaches that combine the full power of statistical and relational learning are at the other end. In this thesis, we present a collection of simple SRL techniques that focus on computational efficiency rather than maximum expressivity, and thereby occupy an intermediate position in the outlined expressivity-efficiency trade-off. The thesis has three main contributions. We first introduce dynamic propositionalization approaches, which provide a simple but principled integration of relational and statistical learners. Dynamic propositionalization is shown to outperform more traditional static propositionalization approaches, while maintaining computational efficiency. A second part presents Markov models for relational sequences, where sequence elements can be logical atoms or complete logical interpretations. By restricting attention to fully observable data and employing a Markov assumption, inference and learning in the resulting formalisms is significantly easier than in more general SRL systems. In a final part, we present two structured probabilistic models that are tailored to particular application domains, namely haplotype reconstruction and activity recognition. These two domains could be modeled using general-purpose statistical relational sequence models; however, the restriction to a particular domain again allows us to derive more efficient special-purpose inference and learning algorithms. The approaches presented throughout the thesis are evaluated in several relational real-world domains, including structure-activity prediction for chemical compounds, web page classification, modeling user behavior in mobile phone networks, and modeling massively multiplayer online games. The field of machine learning is concerned with making predictions based on known examples. For instance, a computer could learn to automatically classify e-mail messages as spam or legitimate messages, based on a number of examples provided by the user. Today, there are a variety of application domains for machine learning. Examples include predicting the biological properties of molecules, applications in information retrieval from the world wide web, or recognizing faces from camera images. Many complex learning tasks have two important properties: 1) the data under consideration is richly structured, such as chemical molecules or the graph structure of the world wide web, and 2) information is often of an uncertain/probabilistic nature. These two issues are addressed in the field of statistical relational learning, which combines relational techniques that can handle structured data with statistical modeling. In this thesis, we present a collection of simple statistical relational learning approaches. Compared to other techniques developed in the field, the resulting systems are slightly less powerful, but much faster and thus applicable to larger data collections. The thesis is organized into three parts, in which we consider different learning settings. In a first part, the goal is to classify individual data items. Techniques developed for this setting are applied to predicting biological properties of molecules. In a second part, we consider modeling sequences of data items. This is relevant for modeling processes that unfold over time, where the state of the process at any given time is structured. As an example domain, we consider modeling the state of massively multiplayer online games, in which a large number of players interact in a structured way. Finally, a third part of the thesis discusses learning approaches that are tailored to particular application domains. By exploiting the specific domain structure, the resulting techniques are significantly faster than general-purpose machine learning tools that could also be applied.

Academic collection --- 681.3*I26 <043> --- 681.3*H <043> --- Learning: analogies; concept learning; induction; knowledge acquisition; language acquisition; parameter learning (Artificial intelligence)--See also {681.3*K32}--Dissertaties --- Information systems--Dissertaties --- Theses --- 681.3*H <043> Information systems--Dissertaties --- 681.3*I26 <043> Learning: analogies; concept learning; induction; knowledge acquisition; language acquisition; parameter learning (Artificial intelligence)--See also {681.3*K32}--Dissertaties