We introduce kLog, a novel language for kernel-based learning on expressive logical and relational representations. kLog allows users to specify logical and relational learning problems declaratively. It builds on simple but powerful concepts: learning from interpretations, entity/relationship data modeling, and logic programming. Access by the kernel to the rich representation is mediated by a technique we call graphicalization: the relational representation is first transformed into a graph — in particular, a grounded entity/relationship diagram. Subsequently, a choice of graph kernel defines the feature space. The kLog framework can be applied to tackle the same range of tasks that has made statistical relational learning so popular, including classification, regression, multitask learning, and collective classification. An empirical evaluation shows that kLog can be either more accurate, or much faster at the same level of accuracy, than Tilde and Alchemy.

Abstract argumentation frameworks (AFs) are one of the central formalisms in AI; equipped with a wide range of semantics, they have proven useful in several application domains. In the thesis we want to complete and extend the recent study on expressiveness of argumentation semantics and use these and other theoretical results for implementations of reasoning tasks in AFs. Moreover, we plan to utilize results on realizability in dynamic scenarios of abstract argumentation, such as revision of argumentation frameworks. Hereby, the knowledge of which extensions can occur together is of central interest when trying to achieve a certain outcome. In other words, the ultimate goal of the thesis is to gain theoretical insights on argumentation semantics in order to employ them in practically efficient reasoning systems for both the evaluation and evolution of AFs.

We design and analyze deterministic truthful approximation mechanisms for multi-unit combinatorial auctions involving a constant number of distinct goods, each in arbitrary limited supply. Prospective buyers (bidders) have preferences over multisets of items, i.e., for more than one unit per distinct good, that are expressed through their private valuation functions. Our objective is to determine allocations of multisets that maximize the Social Welfare approximately. Despite the recent theoretical advances on the design of truthful combinatorial auctions (for multiple distinct goods in unit supply) and multi-unit auctions (for multiple units of a single good), results for the combined setting are much scarcer. We elaborate on the main developments of [Krysta et al., AAMAS 2013], concerning bidders with multi-minded and submodular valuation functions, with an emphasis on the presentation of the relevant algorithmic techniques.

Many electoral control and manipulation problems — which we will refer to in general as manipulative actions problems — are NP-hard in the general case.  Many of these problems fall into polynomial time if the electorate is single-peaked, i.e., is  polarized along some axis/issue. However, real-world electorates are not truly single-peaked — for example, there may be some maverick voters — and to take this into account, we study the complexity of manipulative-action algorithms for  the case of nearly single-peaked electorates.

In this dissertation, we propose an online learning technique to predict a value of a continuous variable by (i) integrating a set of data streams from heterogeneous sources with time varying compositions including (a) changing the quality of data streams, (b) addition or deletion of data streams (ii) integrating the results of several analysis algorithms for each data source when the most suitable algorithm for a given data source is not known a priori (iii) dynamically weighting the prediction of each analysis algorithm and data source on the system prediction based on their varying quality.

Most logic-based machine learning algorithms rely on an Occamist bias where textual simplicity of hypotheses is optimised. This approach, however, fails to distinguish between the efficiencies of hypothesised programs, such as quick sort (O(n log n)) and bubble sort (O(n^2)). We address this issue by considering techniques to minimise both the resource complexity and textual complexity of hypothesised programs. We describe an algorithm proven to learn optimal resource complexity robot strategies, and we propose future work to generalise this approach to a broader class of logic programs.

In this work, we introduce the first approach to the Inverse Reinforcement Learning (IRL) problem in relational domains. IRL has been used to recover a more compact representation of the expert policy leading to better generalization performances among different contexts. On the other hand, relational learning allows representing problems with a varying number of objects (potentially infinite), thus provides more generalizable representations of problems and skills. We show how these different formalisms allow one to create a new IRL algorithm for relational domains that can recover with great efficiency rewards from expert data that have strong generalization and transfer properties. We evaluate our algorithm in representative tasks and study the impact of diverse experimental conditions such as : the number of demonstrations, knowledge about the dynamics, transfer among varying dimensions of a problem, and changing dynamics.

Recently significant advances have been witnessed in the area of distributed word representations based on neural networks, which are also known as word embeddings. Among the new word embedding models, skip-gram negative sampling (SGNS) in the word2vec toolbox has attracted much attention due to its simplicity and effectiveness. However, the principles of SGNS remain not well understood, except for a recent work that explains SGNS as an implicit matrix factorization of the pointwise mutual information (PMI) matrix. In this paper, we provide a new perspective for further understanding SGNS. We point out that SGNS is essentially a representation learning method, which learns to represent the co-occurrence vector for a word. Based on the representation learning view, SGNS is in fact an explicit matrix factorization (EMF) of the words’ co-occurrence matrix. Furthermore, extended supervised word embedding can be established based on our proposed representation learning view.

But growing up, in conflict with her father We present a graphic novel project aiming at illustrating current research results and issues regarding the creative process and its relation with artificial intelligence. The main character, Max Order, is an artist who symbolizes the difficulty of coming up with new, creative ideas, giving up imitation of others and finding one's own style.

Stream reasoning is the task of continuously deriving conclusions on streaming data. As a research theme, it is targeted by different communities which emphasize different aspects, e.g., throughput vs. expressiveness. This thesis aims to advance the theoretical foundations underlying diverse stream reasoning approaches and to convert obtained insights into a prototypical expressive rule-based reasoning system that is lacking to date.