Dynamic Software Product Lines (DSPLs) is still an incipient paradigm to support runtime reconfiguration of systems that need to change their behavior under varying circumstances. As many modern systems demand reconfiguration after post-deployment time, this 10th edition of the DSPL workshops attempts to bring new insight and approaches for such needs.

In computer vision, object reconstruction is the task of inferring the 3D shape of an object based on a single or multiple 2D images. For such purpose, most common frameworks use voxel grids and point clouds. However, both of these approaches have strong limitations. On one hand, the computational cost of using voxels grows cubically as the resolution of the voxels increases. Therefore, 3D object reconstructions are usually set to low resolution. On the other hand, point clouds are unstructured in nature and the proper definition of surfaces and contours is complex. In this study, 3D object reconstruction is carried out applying free-form deformations on pre-existent 3D meshes, through two basic learning processes: template selection and template deformation. From this approach, it is possible to generate high-quality 3D object reconstructions with a lower computational cost. Concretely, two novel lightweight CNNs models are developed and tested: a multi-target learner (Model A) and depth information learner (Model B). According to the results, the performance of the multi-target learner regarding the template selection was around three times better (lower error) than in the baseline architecture, which improved the quality of the 3D reconstructions, whereas the depth-information learner showed promising results in the reconstruction of objects with complex geometry. The inherent issue of using chamfer distance as a loss measure is also examined.

More than fourty participants attended the 9th Symposium on Software Performance in Hildesheim. The "Symposium on Software Performance" brings together researchers and practitioners interested in all facets of software performance, ranging from modeling and prediction to monitoring and runtime management.

Solving a pickup and delivery problem with, e. g., multiple depots, time windows, and heterogeneous vehicles is a challenging routing task. Due to the complexity, a meta-heuristic approach (e. g., a genetic algorithm) with sufficiently good solution quality is recommended. Genetic algorithms contain multiple operators such as the crossover and mutation operators that are called with certain probabilities. However, selecting appropriate probability values (parameters) for these operators strongly depend on the data structure of the given instances. For each new instance, the best parameter configuration must be found to enhance the overall solution quality. In this paper, an a-priori parameter selection approach based on classifying new instances to clusters is presented. Beforehand, a bayesian optimization approach with gaussian processes is used to find the best parameters for each cluster. The a-priori parameter selection is evaluated on four well-known pickup and delivery problem data sets, each with 60 instances and different number of depots.

This paper aims at a cross-lingual analysis of coreference to abstract entities in Czech and German, two languages that are typologi- cally not very close, since they belong to two different language groups – Slavic and Ger- manic. We will specifically focus on coref- erence chains to abstract entities, i.e. verbal phrases, clauses, sentences or even longer text passages. To our knowledge, this type of re- lation is underinvestigated in the current state- of-the-art literature.

The present paper describes an attempt to create an interoperable scheme using existing annotations of textual phenomena across languages and genres including non-canonical ones. Such a kind of analysis requires annotated multilingual resources which are costly. Therefore, we make use of annotations already available in the resources for English, German and Czech. As the annotations in these corpora are based on different conceptual and methodological backgrounds, we need an interoperable scheme that covers existing categories and at the same time allows a comparison of the resources. In this paper, we describe how this interoperable scheme was created and which problematic cases we had to consider. The resulting scheme is supposed to be applied in the future to explore contrasts between the three languages under analysis, for which we expect the greatest differences in the degree of variation between non-canonical and canonical language.

Recommender systems help Web users to address information overload. Their performance, however, depends on the amount of information that users provide about their preferences. Users are not willing to provide information for a large amount of items, thus the quality of recommendations is affected specially for new users. Active learning has been proposed in the past, to acquire preference information from users. Based on an underlying prediction model, these approaches determine the most informative item for querying the new user to provide a rating. In this paper, we propose a new active learning method which is developed specially based on aspect model features. There is a difference between classic active learning and active learning for recommender system. In the recommender system context, each item has already been rated by training users while in classic active learning there is not training user. We take into account this difference and develop a new method which competes with a complicated bayesian approach in accuracy while results in drastically reduced (one order of magnitude) user waiting times, i.e., the time that the users wait before being asked a new query.

Suggesting tasks and learning resources of appropriate difficulty to learners is challenging. Neither should they be too difficult and nor too easy. Well-chosen tasks would enable a quick assessment of the learner, well-chosen learning resources would speed up the learning curve most. We connect active learning to classical pedagogical theory and propose the uncertainty sampling framework as a means to the challenge of selecting optimal tasks and learning resources to learners. To assess the efficiency of this strategy, we compared different exercise selection strategies and evaluated their effect on different datasets. We consistently find that uncertainty sampling significantly outperforms several alternative exercise selection approaches and thus leads to a faster convergence to the true assessment. These findings demonstrate that active (machine) learning is consistent with classic learning theory. It is a valuable instrument for choosing appropriate exercises as well as learning resources both from a teacher’s and from a learner’s perspective.

Selecting promising queries is the key to effective active learning. In this paper, we investigate selection techniques for the task of learning an equivalence relation where the queries are about pairs of objects. As the target relation satisfies the axioms of transitivity, from one queried pair additional constraints can be inferred. We derive both the upper and lower bound on the number of queries needed to converge to the optimal solution. Besides restricting the set of possible solutions, constraints can be used as training data for learning a similarity measure. For selecting queries that result in a large number of meaningful constraints, we present an approximative optimal selection technique that greedily minimizes the expected loss in each round of active learning. This technique makes use of inference of expected constraints. Besides the theoretical results, an extensive evaluation for the application of record linkage shows empirically that the proposed selection method leads to both interesting and a high number of constraints.

We consider the problem to learn a concept or a query in the presence of an ontology formulated in the description logic ELr, in Angluin's framework of active learning that allows the learning algorithm to interactively query an oracle (such as a domain expert). We show that the following can be learned in polynomial time: (1) EL-concepts, (2) symmetry-free ELI-concepts, and (3) conjunctive queries (CQs) that are chordal, symmetry-free, and of bounded arity. In all cases, the learner can pose to the oracle membership queries based on ABoxes and equivalence queries that ask whether a given concept/query from the considered class is equivalent to the target. The restriction to bounded arity in (3) can be removed when we admit unrestricted CQs in equivalence queries. We also show that EL-concepts are not polynomial query learnable in the presence of ELI-ontologies.

We show that ELI queries (ELIQs) are learnable in polynomial time in the presence of a DL-Lite ontology O, in Angluin’s framework of active learning. When initially provided with a conjunctive query (CQ) that implies the target ELIQ under O (in the sense of query containment), it suffices for the learner to only pose membership queries to the oracle, but no equivalence queries. The initial CQ can be obtained by a single equivalence query and is available ‘for free’ in case that O does not pose any disjointness constraints on concepts. Our main technical result is that every ELI concept has only polynomially many most specific subsumers w.r.t. a DL-Lite ontology, generalizing a recent result about homomorphism frontiers by ten Cate and Dalmau.

[Context and motivation] Current and future systems have to operate in complex and dynamic environments. An adaptive system addresses these challenges as it monitors its environment and reacts by changing its behavior. [Question/Problem] Representations of adaptive requirements (e.g., at runtime) and strategies for decision-making have gained a lot of interest in past and current research. Yet, there is a lack of support for elicitation of requirements and environmental information for adaptive systems. [Principal ideas/results] We suggest to apply creativity techniques to elicit adaptation requirements and make use of situations to negotiate them (a situation represents the state of the system and its environment at a particular instance of time). [Contributions] In this paper, we introduce AdaptationExplore, a process for the development of adaptive systems, which supports engineers in particular during the early phases. The results of a pilot study are reported. 37 Master students applied the process on different cases. The study provides first positive experiences on the effectiveness and applicability of the process.

In the latest decades, machine learning approaches have been intensively experimented for natural language processing. Most of the time, systems rely on using statistics within the system, by analyzing texts at the token level and, for labelling tasks, categorizing each among possible classes. One may notice that previous symbolic approaches (e.g. transducers) where designed to delimit pieces of text. Our research team developped mXS, a system that aims at combining both approaches. It locates boundaries of entities by using sequential pattern mining and machine learning. This system, intially developped for French, has been adapted to German.

There is a current trend towards natural, adaptive in-vehicle spoken dialogue systems which react flexibly to the individual requirements of a driver or the driving situation. They aim to provide the driver with the most efficient form of interaction and to thereby reduce his cognitive load. Studies show that even the syntactic form of system output has an influence on the driver and his driving performance. Against this background, in this paper we present our user-centered approach for a user- and situation-adaptive strategy for the syntactic design of voice output. Based on the collected data of two user studies, we combine the two aspects of speech production and perception and compare actual language behaviour with syntactic preferences. The resulting strategy will be evaluated and elaborate further in future user studies.