004 Datenverarbeitung; Informatik
Refine
Year of publication
- 2012 (22) (remove)
Document Type
- Monograph/Edited Volume (11)
- Doctoral Thesis (7)
- Postprint (3)
- Article (1)
Keywords
- AUTOSAR (2)
- Data Integration (2)
- Datenintegration (2)
- Machine Learning (2)
- Maschinelles Lernen (2)
- Model Synchronisation (2)
- Model Transformation (2)
- Modellierung (2)
- SysML (2)
- Tripel-Graph-Grammatik (2)
The field of machine learning studies algorithms that infer predictive models from data. Predictive models are applicable for many practical tasks such as spam filtering, face and handwritten digit recognition, and personalized product recommendation. In general, they are used to predict a target label for a given data instance. In order to make an informed decision about the deployment of a predictive model, it is crucial to know the model’s approximate performance. To evaluate performance, a set of labeled test instances is required that is drawn from the distribution the model will be exposed to at application time. In many practical scenarios, unlabeled test instances are readily available, but the process of labeling them can be a time- and cost-intensive task and may involve a human expert. This thesis addresses the problem of evaluating a given predictive model accurately with minimal labeling effort. We study an active model evaluation process that selects certain instances of the data according to an instrumental sampling distribution and queries their labels. We derive sampling distributions that minimize estimation error with respect to different performance measures such as error rate, mean squared error, and F-measures. An analysis of the distribution that governs the estimator leads to confidence intervals, which indicate how precise the error estimation is. Labeling costs may vary across different instances depending on certain characteristics of the data. For instance, documents differ in their length, comprehensibility, and technical requirements; these attributes affect the time a human labeler needs to judge relevance or to assign topics. To address this, the sampling distribution is extended to incorporate instance-specific costs. We empirically study conditions under which the active evaluation processes are more accurate than a standard estimate that draws equally many instances from the test distribution. We also address the problem of comparing the risks of two predictive models. The standard approach would be to draw instances according to the test distribution, label the selected instances, and apply statistical tests to identify significant differences. Drawing instances according to an instrumental distribution affects the power of a statistical test. We derive a sampling procedure that maximizes test power when used to select instances, and thereby minimizes the likelihood of choosing the inferior model. Furthermore, we investigate the task of comparing several alternative models; the objective of an evaluation could be to rank the models according to the risk that they incur or to identify the model with lowest risk. An experimental study shows that the active procedure leads to higher test power than the standard test in many application domains. Finally, we study the problem of evaluating the performance of ranking functions, which are used for example for web search. In practice, ranking performance is estimated by applying a given ranking model to a representative set of test queries and manually assessing the relevance of all retrieved items for each query. We apply the concepts of active evaluation and active comparison to ranking functions and derive optimal sampling distributions for the commonly used performance measures Discounted Cumulative Gain and Expected Reciprocal Rank. Experiments on web search engine data illustrate significant reductions in labeling costs.
Duplicate detection is the task of identifying all groups of records within a data set that represent the same real-world entity, respectively. This task is difficult, because (i) representations might differ slightly, so some similarity measure must be defined to compare pairs of records and (ii) data sets might have a high volume making a pair-wise comparison of all records infeasible. To tackle the second problem, many algorithms have been suggested that partition the data set and compare all record pairs only within each partition. One well-known such approach is the Sorted Neighborhood Method (SNM), which sorts the data according to some key and then advances a window over the data comparing only records that appear within the same window. We propose several variations of SNM that have in common a varying window size and advancement. The general intuition of such adaptive windows is that there might be regions of high similarity suggesting a larger window size and regions of lower similarity suggesting a smaller window size. We propose and thoroughly evaluate several adaption strategies, some of which are provably better than the original SNM in terms of efficiency (same results with fewer comparisons).
ASP modulo CSP
(2012)
We present the hybrid ASP solver clingcon, combining the simple modeling language and the high performance Boolean solving capacities of Answer Set Programming (ASP) with techniques for using non-Boolean constraints from the area of Constraint Programming (CP). The new clingcon system features an extended syntax supporting global constraints and optimize statements for constraint variables. The major technical innovation improves the interaction between ASP and CP solver through elaborated learning techniques based on irreducible inconsistent sets. A broad empirical evaluation shows that these techniques yield a performance improvement of an order of magnitude.
This document presents an axiom selection technique for classic first order theorem proving based on the relevance of axioms for the proof of a conjecture. It is based on unifiability of predicates and does not need statistical information like symbol frequency. The scope of the technique is the reduction of the set of axioms and the increase of the amount of provable conjectures in a given time. Since the technique generates a subset of the axiom set, it can be used as a preprocessor for automated theorem proving. This technical report describes the conception, implementation and evaluation of ARDE. The selection method, which is based on a breadth-first graph search by unifiability of predicates, is a weakened form of the connection calculus and uses specialised variants or unifiability to speed up the selection. The implementation of the concept is evaluated with comparison to the results of the world championship of theorem provers of the year 2012 (CASC J6). It is shown that both the theorem prover leanCoP which uses the connection calculus and E which uses equality reasoning, can benefit from the selection approach. Also, the evaluation shows that the concept is applyable for theorem proving problems with thousands of formulae and that the selection is independent from the calculus used by the theorem prover.
Data dependencies, or integrity constraints, are used to improve the quality of a database schema, to optimize queries, and to ensure consistency in a database. In the last years conditional dependencies have been introduced to analyze and improve data quality. In short, a conditional dependency is a dependency with a limited scope defined by conditions over one or more attributes. Only the matching part of the instance must adhere to the dependency. In this paper we focus on conditional inclusion dependencies (CINDs). We generalize the definition of CINDs, distinguishing covering and completeness conditions. We present a new use case for such CINDs showing their value for solving complex data quality tasks. Further, we define quality measures for conditions inspired by precision and recall. We propose efficient algorithms that identify covering and completeness conditions conforming to given quality thresholds. Our algorithms choose not only the condition values but also the condition attributes automatically. Finally, we show that our approach efficiently provides meaningful and helpful results for our use case.
Cyber-physical systems achieve sophisticated system behavior exploring the tight interconnection of physical coupling present in classical engineering systems and information technology based coupling. A particular challenging case are systems where these cyber-physical systems are formed ad hoc according to the specific local topology, the available networking capabilities, and the goals and constraints of the subsystems captured by the information processing part. In this paper we present a formalism that permits to model the sketched class of cyber-physical systems. The ad hoc formation of tightly coupled subsystems of arbitrary size are specified using a UML-based graph transformation system approach. Differential equations are employed to define the resulting tightly coupled behavior. Together, both form hybrid graph transformation systems where the graph transformation rules define the discrete steps where the topology or modes may change, while the differential equations capture the continuous behavior in between such discrete changes. In addition, we demonstrate that automated analysis techniques known for timed graph transformation systems for inductive invariants can be extended to also cover the hybrid case for an expressive case of hybrid models where the formed tightly coupled subsystems are restricted to smaller local networks.
Fußverkehr findet im gesamten öffentlichen Raum statt und ermöglicht die lückenlose Verbindung von Tür zu Tür. Jeder Mensch steht vor Beginn einer Fortbewegung vor den Fragen „Wo bin ich?“, „Wo liegt mein Ziel?“ und „Wie komme ich dahin?“. Ein Großteil der auf dem Markt befindlichen Navigationssysteme für Fußgänger stellen reduzierte Varianten aus Fahrzeugen dar und basieren auf 2D- Kartendarstellungen oder bilden die Realität als dreidimensionales Modell ab. Navigationsprobleme entstehen dann, wenn es dem Nutzer nicht gelingt, die Information aus der Anweisung auf die Wirklichkeit zu beziehen und umzusetzen. Ein möglicher Grund dafür liegt in der Visualisierung der Navigationsanweisung. Die räumliche Wahrnehmung des Menschen erfolgt ausgehend von einem bestimmten Betrachtungsstandpunkt und bringt die Lage von Objekten und deren Beziehung zueinander zum Ausdruck. Der Einsatz von Augmented Reality (erweiterte Realität) entspricht dem Erscheinungsbild der menschlichen Wahrnehmung und ist für Menschen eine natürliche und zugleich vertraute Ansichtsform. Im Unterschied zu kartographischer Visualisierung wird die Umwelt mittels Augmented Reality nicht modelliert, sondern realitätsgetreu abgebildet und ergänzt. Das Ziel dieser Arbeit ist ein Navigationsverfahren, das der natürlichen Fort-bewegung und Sichtweise von Fußgängern gerecht wird. Das Konzept basiert auf dem Einsatz einer Kombination aus Realität und virtueller Realität zu einer erweiterten Ansicht. Da keine Darstellungsform als die Route selbst besser geeignet ist, um einen Routenverlauf zu beschreiben, wird die Realität durch eine virtuelle Route erweitert. Die perspektivische Anpassung der Routendarstellung erfordert die sensorische Erfassung der Position und Lage des Betrachtungsstandpunktes. Das der Navigation zu Grunde liegende Datenmodell bleibt dem Betrachter dabei verborgen und ist nur in Form der erweiterten Realität sichtbar. Der im Rahmen dieser Arbeit entwickelte Prototyp trägt die Bezeichnung RealityView. Die Basis bildet ein freies und quelloffenes Navigationssystem, das für die Fußgängernavigation modular erweitert wurde. Das Ergebnis ist ein smartphonebasierter Navigationsprototyp, in dem die Ansichtsform einer zweidimensionalen Bildschirmkarte im Grundriss und die Darstellung einer erweiterten Realität im Aufriss kombiniert werden. Die Evaluation des Prototyps bestätigt die Hypothese, dass der Einsatz von Augmented Reality für die Navigation von Fußgängern möglich ist und von der Nutzergruppe akzeptiert wird. Darüber hinaus bescheinigen Wissenschaftler im Rahmen von Experten-interviews den konzeptionellen Ansatz und die prototypische Umsetzung des RealityView. Die Auswertung einer Eye-Tracking-Pilotstudie erbrachte den Nachweis, dass Fußgänger die Navigationsanweisung auf markante Objekte der Umwelt beziehen, deren Auswahl durch den Einsatz von Augmented Reality begünstigt wird.
F2C2
(2012)
Background: Flux coupling analysis (FCA) has become a useful tool in the constraint-based analysis of genome-scale metabolic networks. FCA allows detecting dependencies between reaction fluxes of metabolic networks at steady-state. On the one hand, this can help in the curation of reconstructed metabolic networks by verifying whether the coupling between reactions is in agreement with the experimental findings. On the other hand, FCA can aid in defining intervention strategies to knock out target reactions.
Results: We present a new method F2C2 for FCA, which is orders of magnitude faster than previous approaches. As a consequence, FCA of genome-scale metabolic networks can now be performed in a routine manner.
Conclusions: We propose F2C2 as a fast tool for the computation of flux coupling in genome-scale metabolic networks. F2C2 is freely available for non-commercial use at https://sourceforge.net/projects/f2c2/files/.
In-Memory Data Management
(2012)
Nach 50 Jahren erfolgreicher Entwicklunghat die Business-IT einen neuenWendepunkt erreicht. Hier zeigen die Autoren erstmalig, wieIn-Memory Computing dieUnternehmensprozesse künftig verändern wird. Bisher wurden Unternehmensdaten aus Performance-Gründen auf verschiedene Datenbanken verteilt: Analytische Datenresidieren in Data Warehouses und werden regelmäßig mithilfe transaktionaler Systeme synchronisiert. Diese Aufspaltung macht flexibles Echtzeit-Reporting aktueller Daten unmöglich. Doch dank leistungsfähigerMulti-Core-CPUs, großer Hauptspeicher, Cloud Computing und immerbesserer mobiler Endgeräte lassen die Unternehmen dieses restriktive Modell zunehmend hinter sich. Die Autoren stellen Techniken vor, die eine analytische und transaktionale Verarbeitung in Echtzeit erlauben und so dem Geschäftsleben neue Wege bahnen.
During the overall development of complex engineering systems different modeling notations are employed. For example, in the domain of automotive systems system engineering models are employed quite early to capture the requirements and basic structuring of the entire system, while software engineering models are used later on to describe the concrete software architecture. Each model helps in addressing the specific design issue with appropriate notations and at a suitable level of abstraction. However, when we step forward from system design to the software design, the engineers have to ensure that all decisions captured in the system design model are correctly transferred to the software engineering model. Even worse, when changes occur later on in either model, today the consistency has to be reestablished in a cumbersome manual step. In this report, we present in an extended version of [Holger Giese, Stefan Neumann, and Stephan Hildebrandt. Model Synchronization at Work: Keeping SysML and AUTOSAR Models Consistent. In Gregor Engels, Claus Lewerentz, Wilhelm Schäfer, Andy Schürr, and B. Westfechtel, editors, Graph Transformations and Model Driven Enginering - Essays Dedicated to Manfred Nagl on the Occasion of his 65th Birthday, volume 5765 of Lecture Notes in Computer Science, pages 555–579. Springer Berlin / Heidelberg, 2010.] how model synchronization and consistency rules can be applied to automate this task and ensure that the different models are kept consistent. We also introduce a general approach for model synchronization. Besides synchronization, the approach consists of tool adapters as well as consistency rules covering the overlap between the synchronized parts of a model and the rest. We present the model synchronization algorithm based on triple graph grammars in detail and further exemplify the general approach by means of a model synchronization solution between system engineering models in SysML and software engineering models in AUTOSAR which has been developed for an industrial partner. In the appendix as extension to [19] the meta-models and all TGG rules for the SysML to AUTOSAR model synchronization are documented.