004 Datenverarbeitung; Informatik
Refine
Has Fulltext
- yes (108) (remove)
Year of publication
Document Type
- Doctoral Thesis (51)
- Article (40)
- Postprint (6)
- Conference Proceeding (5)
- Master's Thesis (3)
- Bachelor Thesis (1)
- Habilitation Thesis (1)
- Preprint (1)
Language
- English (108) (remove)
Keywords
- Maschinelles Lernen (7)
- Computer Science Education (5)
- Machine Learning (5)
- Antwortmengenprogrammierung (4)
- answer set programming (4)
- Answer Set Programming (3)
- Competence Measurement (3)
- DPLL (3)
- Secondary Education (3)
- Algorithmen (2)
Institute
- Institut für Informatik und Computational Science (108) (remove)
In this bachelor’s thesis I implement the automatic theorem prover nanoCoP-Ω. This system is the result of porting arithmetic and equality handling procedures first introduced in the automatic theorem prover with arithmetic leanCoP-Ω into the similar system nanoCoP 2.0. To understand these procedures, I first introduce the mathematical background to both automatic theorem proving and arithmetic expressions. I present the predecessor projects leanCoP, nanoCoP and leanCoP-Ω, out of which nanCoP-Ω was developed. This is followed by an extensive description of the concepts the non-clausal connection calculus needed to be extended by, to allow for proving arithmetic expressions and equalities, as well as of their implementation into nanoCoP-Ω. An extensive comparison between both the runtimes and the number of solved problems of the systems nanoCoP-Ω and leanCoP-Ω was made. I come to the conclusion, that nanoCoP-Ω is considerably faster than leanCoP-Ω for small problems, though less well suited for larger problems. Additionally, I was able to construct a non-theorem that nanoCoP-Ω generates a false proof for. I discuss how this pressing issue could be resolved, as well as some possible optimizations and expansions of the system.
Accurately solving classification problems nowadays is likely to be the most relevant machine learning task. Binary classification separating two classes only is algorithmically simpler but has fewer potential applications as many real-world problems are multi-class. On the reverse, separating only a subset of classes simplifies the classification task. Even though existing multi-class machine learning algorithms are very flexible regarding the number of classes, they assume that the target set Y is fixed and cannot be restricted once the training is finished. On the other hand, existing state-of-the-art production environments are becoming increasingly interconnected with the advance of Industry 4.0 and related technologies such that additional information can simplify the respective classification problems. In light of this, the main aim of this thesis is to introduce dynamic classification that generalizes multi-class classification such that the target class set can be restricted arbitrarily to a non-empty class subset M of Y at any time between two consecutive predictions.
This task is solved by a combination of two algorithmic approaches. First, classifier calibration, which transforms predictions into posterior probability estimates that are intended to be well calibrated. The analysis provided focuses on monotonic calibration and in particular corrects wrong statements that appeared in the literature. It also reveals that bin-based evaluation metrics, which became popular in recent years, are unjustified and should not be used at all. Next, the validity of Platt scaling, which is the most relevant parametric calibration approach, is analyzed in depth. In particular, its optimality for classifier predictions distributed according to four different families of probability distributions as well its equivalence with Beta calibration up to a sigmoidal preprocessing are proven. For non-monotonic calibration, extended variants on kernel density estimation and the ensemble method EKDE are introduced. Finally, the calibration techniques are evaluated using a simulation study with complete information as well as on a selection of 46 real-world data sets.
Building on this, classifier calibration is applied as part of decomposition-based classification that aims to reduce multi-class problems to simpler (usually binary) prediction tasks. For the involved fusing step performed at prediction time, a new approach based on evidence theory is presented that uses classifier calibration to model mass functions. This allows the analysis of decomposition-based classification against a strictly formal background and to prove closed-form equations for the overall combinations. Furthermore, the same formalism leads to a consistent integration of dynamic class information, yielding a theoretically justified and computationally tractable dynamic classification model. The insights gained from this modeling are combined with pairwise coupling, which is one of the most relevant reduction-based classification approaches, such that all individual predictions are combined with a weight. This not only generalizes existing works on pairwise coupling but also enables the integration of dynamic class information.
Lastly, a thorough empirical study is performed that compares all newly introduced approaches to existing state-of-the-art techniques. For this, evaluation metrics for dynamic classification are introduced that depend on corresponding sampling strategies. Thereafter, these are applied during a three-part evaluation. First, support vector machines and random forests are applied on 26 data sets from the UCI Machine Learning Repository. Second, two state-of-the-art deep neural networks are evaluated on five benchmark data sets from a relatively recent reference work. Here, computationally feasible strategies to apply the presented algorithms in combination with large-scale models are particularly relevant because a naive application is computationally intractable. Finally, reference data from a real-world process allowing the inclusion of dynamic class information are collected and evaluated. The results show that in combination with support vector machines and random forests, pairwise coupling approaches yield the best results, while in combination with deep neural networks, differences between the different approaches are mostly small to negligible. Most importantly, all results empirically confirm that dynamic classification succeeds in improving the respective prediction accuracies. Therefore, it is crucial to pass dynamic class information in respective applications, which requires an appropriate digital infrastructure.
The highly structured nature of the educational sector demands effective policy mechanisms close to the needs of the field. That is why evidence-based policy making, endorsed by the European Commission under Erasmus+ Key Action 3, aims to make an alignment between the domains of policy and practice. Against this background, this article addresses two issues: First, that there is a vertical gap in the translation of higher-level policies to local strategies and regulations. Second, that there is a horizontal gap between educational domains regarding the policy awareness of individual players. This was analyzed in quantitative and qualitative studies with domain experts from the fields of virtual mobility and teacher training. From our findings, we argue that the combination of both gaps puts the academic bridge from secondary to tertiary education at risk, including the associated knowledge proficiency levels. We discuss the role of digitalization in the academic bridge by asking the question: which value does the involved stakeholders expect from educational policies? As a theoretical basis, we rely on the model of value co-creation for and by stakeholders. We describe the used instruments along with the obtained results and proposed benefits. Moreover, we reflect on the methodology applied, and we finally derive recommendations for future academic bridge policies.
Research publications and data nowadays should be publicly available on the internet and, theoretically, usable for everyone to develop further research, products, or services. The long-term accessibility of research data is, therefore, fundamental in the economy of the research production process. However, the availability of data is not sufficient by itself, but also their quality must be verifiable. Measures to ensure reuse and reproducibility need to include the entire research life cycle, from the experimental design to the generation of data, quality control, statistical analysis, interpretation, and validation of the results. Hence, high-quality records, particularly for providing a string of documents for the verifiable origin of data, are essential elements that can act as a certificate for potential users (customers). These records also improve the traceability and transparency of data and processes, therefore, improving the reliability of results. Standards for data acquisition, analysis, and documentation have been fostered in the last decade driven by grassroot initiatives of researchers and organizations such as the Research Data Alliance (RDA). Nevertheless, what is still largely missing in the life science academic research are agreed procedures for complex routine research workflows. Here, well-crafted documentation like standard operating procedures (SOPs) offer clear direction and instructions specifically designed to avoid deviations as an absolute necessity for reproducibility. Therefore, this paper provides a standardized workflow that explains step by step how to write an SOP to be used as a starting point for appropriate research documentation.
In the last decades, there was a notable progress in solving the well-known Boolean satisfiability (Sat) problem, which can be witnessed by powerful Sat solvers. One of the reasons why these solvers are so fast are structural properties of instances that are utilized by the solver’s interna. This thesis deals with the well-studied structural property treewidth, which measures the closeness of an instance to being a tree. In fact, there are many problems parameterized by treewidth that are solvable in polynomial time in the instance size when parameterized by treewidth.
In this work, we study advanced treewidth-based methods and tools for problems in knowledge representation and reasoning (KR). Thereby, we provide means to establish precise runtime results (upper bounds) for canonical problems relevant to KR. Then, we present a new type of problem reduction, which we call decomposition-guided (DG) that
allows us to precisely monitor the treewidth when reducing from one problem to another problem. This new reduction type will be the basis for a long-open lower bound result for quantified Boolean formulas and allows us to design a new methodology for establishing runtime lower bounds for problems parameterized by treewidth.
Finally, despite these lower bounds, we provide an efficient implementation of algorithms that adhere to treewidth. Our approach finds suitable abstractions of instances, which are subsequently refined in a recursive fashion, and it uses Sat solvers for solving subproblems. It turns out that our resulting solver is quite competitive for two canonical counting problems related to Sat.
A common feature in Answer Set Programming is the use of a second negation, stronger than default negation and sometimes called explicit, strong or classical negation. This explicit negation is normally used in front of atoms, rather than allowing its use as a regular operator. In this paper we consider the arbitrary combination of explicit negation with nested expressions, as those defined by Lifschitz, Tang and Turner. We extend the concept of reduct for this new syntax and then prove that it can be captured by an extension of Equilibrium Logic with this second negation. We study some properties of this variant and compare to the already known combination of Equilibrium Logic with Nelson's strong negation.
In a recent line of research, two familiar concepts from logic programming semantics (unfounded sets and splitting) were extrapolated to the case of epistemic logic programs. The property of epistemic splitting provides a natural and modular way to understand programs without epistemic cycles but, surprisingly, was only fulfilled by Gelfond's original semantics (G91), among the many proposals in the literature. On the other hand, G91 may suffer from a kind of self-supported, unfounded derivations when epistemic cycles come into play. Recently, the absence of these derivations was also formalised as a property of epistemic semantics called foundedness. Moreover, a first semantics proved to satisfy foundedness was also proposed, the so-called Founded Autoepistemic Equilibrium Logic (FAEEL). In this paper, we prove that FAEEL also satisfies the epistemic splitting property something that, together with foundedness, was not fulfilled by any other approach up to date. To prove this result, we provide an alternative characterisation of FAEEL as a combination of G91 with a simpler logic we called Founded Epistemic Equilibrium Logic (FEEL), which is somehow an extrapolation of the stable model semantics to the modal logic S5.
Parsing of argumentative structures has become a very active line of research in recent years. Like discourse parsing or any other natural language task that requires prediction of linguistic structures, most approaches choose to learn a local model and then perform global decoding over the local probability distributions, often imposing constraints that are specific to the task at hand. Specifically for argumentation parsing, two decoding approaches have been recently proposed: Minimum Spanning Trees (MST) and Integer Linear Programming (ILP), following similar trends in discourse parsing. In contrast to discourse parsing though, where trees are not always used as underlying annotation schemes, argumentation structures so far have always been represented with trees. Using the 'argumentative microtext corpus' [in: Argumentation and Reasoned Action: Proceedings of the 1st European Conference on Argumentation, Lisbon 2015 / Vol. 2, College Publications, London, 2016, pp. 801-815] as underlying data and replicating three different decoding mechanisms, in this paper we propose a novel ILP decoder and an extension to our earlier MST work, and then thoroughly compare the approaches. The result is that our new decoder outperforms related work in important respects, and that in general, ILP and MST yield very similar performance.
Emotions are a central element of human experience. They occur with high frequency in everyday life and play an important role in decision making. However, currently there is no consensus among researchers on what constitutes an emotion and on how emotions should be investigated. This dissertation identifies three problems of current emotion research: the problem of ground truth, the problem of incomplete constructs and the problem of optimal representation. I argue for a focus on the detailed measurement of emotion manifestations with computer-aided methods to solve these problems. This approach is demonstrated in three research projects, which describe the development of methods specific to these problems as well as their application to concrete research questions.
The problem of ground truth describes the practice to presuppose a certain structure of emotions as the a priori ground truth. This determines the range of emotion descriptions and sets a standard for the correct assignment of these descriptions. The first project illustrates how this problem can be circumvented with a multidimensional emotion perception paradigm which stands in contrast to the emotion recognition paradigm typically employed in emotion research. This paradigm allows to calculate an objective difficulty measure and to collect subjective difficulty ratings for the perception of emotional stimuli. Moreover, it enables the use of an arbitrary number of emotion stimuli categories as compared to the commonly used six basic emotion categories. Accordingly, we collected data from 441 participants using dynamic facial expression stimuli from 40 emotion categories. Our findings suggest an increase in emotion perception difficulty with increasing actor age and provide evidence to suggest that young adults, the elderly and men underestimate their emotion perception difficulty. While these effects were predicted from the literature, we also found unexpected and novel results. In particular, the increased difficulty on the objective difficulty measure for female actors and observers stood in contrast to reported findings. Exploratory analyses revealed low relevance of person-specific variables for the prediction of emotion perception difficulty, but highlighted the importance of a general pleasure dimension for the ease of emotion perception.
The second project targets the problem of incomplete constructs which relates to vaguely defined psychological constructs on emotion with insufficient ties to tangible manifestations. The project exemplifies how a modern data collection method such as face tracking data can be used to sharpen these constructs on the example of arousal, a long-standing but fuzzy construct in emotion research. It describes how measures of distance, speed and magnitude of acceleration can be computed from face tracking data and investigates their intercorrelations. We find moderate to strong correlations among all measures of static information on one hand and all measures of dynamic information on the other. The project then investigates how self-rated arousal is tied to these measures in 401 neurotypical individuals and 19 individuals with autism. Distance to the neutral face was predictive of arousal ratings in both groups. Lower mean arousal ratings were found for the autistic group, but no difference in correlation of the measures and arousal ratings could be found between groups. Results were replicated in a high autistic traits group consisting of 41 participants. The findings suggest a qualitatively similar perception of arousal for individuals with and without autism. No correlations between valence ratings and any of the measures could be found which emphasizes the specificity of our tested measures for the construct of arousal.
The problem of optimal representation refers to the search for the best representation of emotions and the assumption that there is a one-fits-all solution. In the third project we introduce partial least squares analysis as a general method to find an optimal representation to relate two high-dimensional data sets to each other. The project demonstrates its applicability to emotion research on the question of emotion perception differences between men and women. The method was used with emotion rating data from 441 participants and face tracking data computed on 306 videos. We found quantitative as well as qualitative differences in the perception of emotional facial expressions between these groups. We showed that women’s emotional perception systematically captured more of the variance in facial expressions. Additionally, we could show that significant differences exist in the way that women and men perceive some facial expressions which could be visualized as concrete facial expression sequences. These expressions suggest differing perceptions of masked and ambiguous facial expressions between the sexes. In order to facilitate use of the developed method by the research community, a package for the statistical environment R was written. Furthermore, to call attention to the method and its usefulness for emotion research, a website was designed that allows users to explore a model of emotion ratings and facial expression data in an interactive fashion.
Services that operate over the Internet are under constant threat of being exposed to fraudulent use. Maintaining good user experience for legitimate users often requires the classification of entities as malicious or legitimate in order to initiate countermeasures. As an example, inbound email spam filters decide for spam or non-spam. They can base their decision on both the content of each email as well as on features that summarize prior emails received from the sending server. In general, discriminative classification methods learn to distinguish positive from negative entities. Each decision for a label may be based on features of the entity and related entities. When labels of related entities have strong interdependencies---as can be assumed e.g. for emails being delivered by the same user---classification decisions should not be made independently and dependencies should be modeled in the decision function. This thesis addresses the formulation of discriminative classification problems that are tailored for the specific demands of the following three Internet security applications. Theoretical and algorithmic solutions are devised to protect an email service against flooding of user inboxes, to mitigate abusive usage of outbound email servers, and to protect web servers against distributed denial of service attacks.
In the application of filtering an inbound email stream for unsolicited emails, utilizing features that go beyond each individual email's content can be valuable. Information about each sending mail server can be aggregated over time and may help in identifying unwanted emails. However, while this information will be available to the deployed email filter, some parts of the training data that are compiled by third party providers may not contain this information. The missing features have to be estimated at training time in order to learn a classification model. In this thesis an algorithm is derived that learns a decision function that integrates over a distribution of values for each missing entry. The distribution of missing values is a free parameter that is optimized to learn an optimal decision function.
The outbound stream of emails of an email service provider can be separated by the customer IDs that ask for delivery. All emails that are sent by the same ID in the same period of time are related, both in content and in label. Hijacked customer accounts may send batches of unsolicited emails to other email providers, which in turn might blacklist the sender's email servers after detection of incoming spam emails. The risk of being blocked from further delivery depends on the rate of outgoing unwanted emails and the duration of high spam sending rates. An optimization problem is developed that minimizes the expected cost for the email provider by learning a decision function that assigns a limit on the sending rate to customers based on the each customer's email stream.
Identifying attacking IPs during HTTP-level DDoS attacks allows to block those IPs from further accessing the web servers. DDoS attacks are usually carried out by infected clients that are members of the same botnet and show similar traffic patterns. HTTP-level attacks aim at exhausting one or more resources of the web server infrastructure, such as CPU time. If the joint set of attackers cannot increase resource usage close to the maximum capacity, no effect will be experienced by legitimate users of hosted web sites. However, if the additional load raises the computational burden towards the critical range, user experience will degrade until service may be unavailable altogether. As the loss of missing one attacker depends on block decisions for other attackers---if most other attackers are detected, not blocking one client will likely not be harmful---a structured output model has to be learned. In this thesis an algorithm is developed that learns a structured prediction decoder that searches the space of label assignments, guided by a policy.
Each model is evaluated on real-world data and is compared to reference methods. The results show that modeling each classification problem according to the specific demands of the task improves performance over solutions that do not consider the constraints inherent to an application.