TY - CHAP AU - Hanselle, Jonas Manuel AU - Hüllermeier, Eyke AU - Mohr, Felix AU - Ngonga Ngomo, Axel-Cyrille AU - Sherif, Mohamed AU - Tornede, Alexander AU - Wever, Marcel Dominik ED - Haake, Claus-Jochen ED - Meyer auf der Heide, Friedhelm ED - Platzner, Marco ED - Wachsmuth, Henning ED - Wehrheim, Heike ID - 45884 T2 - On-The-Fly Computing -- Individualized IT-services in dynamic markets TI - Configuration and Evaluation VL - 412 ER - TY - THES AU - Tornede, Alexander ID - 45780 TI - Advanced Algorithm Selection with Machine Learning: Handling Large Algorithm Sets, Learning From Censored Data, and Simplyfing Meta Level Decisions ER - TY - BOOK AB - In the proposal for our CRC in 2011, we formulated a vision of markets for IT services that describes an approach to the provision of such services that was novel at that time and, to a large extent, remains so today: „Our vision of on-the-fly computing is that of IT services individually and automatically configured and brought to execution from flexibly combinable services traded on markets. At the same time, we aim at organizing markets whose participants maintain a lively market of services through appropriate entrepreneurial actions.“ Over the last 12 years, we have developed methods and techniques to address problems critical to the convenient, efficient, and secure use of on-the-fly computing. Among other things, we have made the description of services more convenient by allowing natural language input, increased the quality of configured services through (natural language) interaction and more efficient configuration processes and analysis procedures, made the quality of (the products of) providers in the marketplace transparent through reputation systems, and increased the resource efficiency of execution through reconfigurable heterogeneous computing nodes and an integrated treatment of service description and configuration. We have also developed network infrastructures that have a high degree of adaptivity, scalability, efficiency, and reliability, and provide cryptographic guarantees of anonymity and security for market participants and their products and services. To demonstrate the pervasiveness of the OTF computing approach, we have implemented a proof-of-concept for OTF computing that can run typical scenarios of an OTF market. We illustrated the approach using a cutting-edge application scenario – automated machine learning (AutoML). Finally, we have been pushing our work for the perpetuation of On-The-Fly Computing beyond the SFB and sharing the expertise gained in the SFB in events with industry partners as well as transfer projects. This work required a broad spectrum of expertise. Computer scientists and economists with research interests such as computer networks and distributed algorithms, security and cryptography, software engineering and verification, configuration and machine learning, computer engineering and HPC, microeconomics and game theory, business informatics and management have successfully collaborated here. AU - Haake, Claus-Jochen AU - Meyer auf der Heide, Friedhelm AU - Platzner, Marco AU - Wachsmuth, Henning AU - Wehrheim, Heike ID - 45863 TI - On-The-Fly Computing -- Individualized IT-services in dynamic markets VL - 412 ER - TY - GEN AB - Algorithm configuration (AC) is concerned with the automated search of the most suitable parameter configuration of a parametrized algorithm. There is currently a wide variety of AC problem variants and methods proposed in the literature. Existing reviews do not take into account all derivatives of the AC problem, nor do they offer a complete classification scheme. To this end, we introduce taxonomies to describe the AC problem and features of configuration methods, respectively. We review existing AC literature within the lens of our taxonomies, outline relevant design choices of configuration approaches, contrast methods and problem variants against each other, and describe the state of AC in industry. Finally, our review provides researchers and practitioners with a look at future research directions in the field of AC. AU - Schede, Elias AU - Brandt, Jasmin AU - Tornede, Alexander AU - Wever, Marcel Dominik AU - Bengs, Viktor AU - Hüllermeier, Eyke AU - Tierney, Kevin ID - 30868 T2 - arXiv:2202.01651 TI - A Survey of Methods for Automated Algorithm Configuration ER - TY - CONF AB - It is well known that different algorithms perform differently well on an instance of an algorithmic problem, motivating algorithm selection (AS): Given an instance of an algorithmic problem, which is the most suitable algorithm to solve it? As such, the AS problem has received considerable attention resulting in various approaches - many of which either solve a regression or ranking problem under the hood. Although both of these formulations yield very natural ways to tackle AS, they have considerable weaknesses. On the one hand, correctly predicting the performance of an algorithm on an instance is a sufficient, but not a necessary condition to produce a correct ranking over algorithms and in particular ranking the best algorithm first. On the other hand, classical ranking approaches often do not account for concrete performance values available in the training data, but only leverage rankings composed from such data. We propose HARRIS- Hybrid rAnking and RegRessIon foreSts - a new algorithm selector leveraging special forests, combining the strengths of both approaches while alleviating their weaknesses. HARRIS' decisions are based on a forest model, whose trees are created based on splits optimized on a hybrid ranking and regression loss function. As our preliminary experimental study on ASLib shows, HARRIS improves over standard algorithm selection approaches on some scenarios showing that combining ranking and regression in trees is indeed promising for AS. AU - Fehring, Lukass AU - Hanselle, Jonas Manuel AU - Tornede, Alexander ID - 34103 T2 - Workshop on Meta-Learning (MetaLearn 2022) @ NeurIPS 2022 TI - HARRIS: Hybrid Ranking and Regression Forests for Algorithm Selection ER - TY - CONF AB - The creation of an RDF knowledge graph for a particular application commonly involves a pipeline of tools that transform a set ofinput data sources into an RDF knowledge graph in a process called dataset augmentation. The components of such augmentation pipelines often require extensive configuration to lead to satisfactory results. Thus, non-experts are often unable to use them. Wepresent an efficient supervised algorithm based on genetic programming for learning knowledge graph augmentation pipelines of arbitrary length. Our approach uses multi-expression learning to learn augmentation pipelines able to achieve a high F-measure on the training data. Our evaluation suggests that our approach can efficiently learn a larger class of RDF dataset augmentation tasks than the state of the art while using only a single training example. Even on the most complex augmentation problem we posed, our approach consistently achieves an average F1-measure of 99% in under 500 iterations with an average runtime of 16 seconds AU - Dreßler, Kevin AU - Sherif, Mohamed AU - Ngonga Ngomo, Axel-Cyrille ID - 31806 KW - 2022 RAKI SFB901 deer dice kevin knowgraphs limes ngonga sherif simba T2 - Proceedings of the 33rd ACM Conference on Hypertext and Hypermedia TI - ADAGIO - Automated Data Augmentation of Knowledge Graphs Using Multi-expression Learning ER - TY - GEN AU - Fehring, Lukas ID - 33033 TI - Combined Ranking and Regression Trees for Algorithm Selection ER - TY - GEN AB - In online algorithm selection (OAS), instances of an algorithmic problem class are presented to an agent one after another, and the agent has to quickly select a presumably best algorithm from a fixed set of candidate algorithms. For decision problems such as satisfiability (SAT), quality typically refers to the algorithm's runtime. As the latter is known to exhibit a heavy-tail distribution, an algorithm is normally stopped when exceeding a predefined upper time limit. As a consequence, machine learning methods used to optimize an algorithm selection strategy in a data-driven manner need to deal with right-censored samples, a problem that has received little attention in the literature so far. In this work, we revisit multi-armed bandit algorithms for OAS and discuss their capability of dealing with the problem. Moreover, we adapt them towards runtime-oriented losses, allowing for partially censored data while keeping a space- and time-complexity independent of the time horizon. In an extensive experimental evaluation on an adapted version of the ASlib benchmark, we demonstrate that theoretically well-founded methods based on Thompson sampling perform specifically strong and improve in comparison to existing methods. AU - Tornede, Alexander AU - Bengs, Viktor AU - Hüllermeier, Eyke ID - 30867 T2 - Proceedings of the 36th AAAI Conference on Artificial Intelligence TI - Machine Learning for Online Algorithm Selection under Censored Feedback ER - TY - GEN AB - The problem of selecting an algorithm that appears most suitable for a specific instance of an algorithmic problem class, such as the Boolean satisfiability problem, is called instance-specific algorithm selection. Over the past decade, the problem has received considerable attention, resulting in a number of different methods for algorithm selection. Although most of these methods are based on machine learning, surprisingly little work has been done on meta learning, that is, on taking advantage of the complementarity of existing algorithm selection methods in order to combine them into a single superior algorithm selector. In this paper, we introduce the problem of meta algorithm selection, which essentially asks for the best way to combine a given set of algorithm selectors. We present a general methodological framework for meta algorithm selection as well as several concrete learning methods as instantiations of this framework, essentially combining ideas of meta learning and ensemble learning. In an extensive experimental evaluation, we demonstrate that ensembles of algorithm selectors can significantly outperform single algorithm selectors and have the potential to form the new state of the art in algorithm selection. AU - Tornede, Alexander AU - Gehring, Lukas AU - Tornede, Tanja AU - Wever, Marcel Dominik AU - Hüllermeier, Eyke ID - 30865 T2 - Machine Learning TI - Algorithm Selection on a Meta Level ER - TY - JOUR AB - AbstractHeated tool butt welding is a method often used for joining thermoplastics, especially when the components are made out of different materials. The quality of the connection between the components crucially depends on a suitable choice of the parameters of the welding process, such as heating time, temperature, and the precise way how the parts are then welded. Moreover, when different materials are to be joined, the parameter values need to be tailored to the specifics of the respective material. To this end, in this paper, three approaches to tailor the parameter values to optimize the quality of the connection are compared: a heuristic by Potente, statistical experimental design, and Bayesian optimization. With the suitability for practice in mind, a series of experiments are carried out with these approaches, and their capabilities of proposing well-performing parameter values are investigated. As a result, Bayesian optimization is found to yield peak performance, but the costs for optimization are substantial. In contrast, the Potente heuristic does not require any experimentation and recommends parameter values with competitive quality. AU - Gevers, Karina AU - Tornede, Alexander AU - Wever, Marcel Dominik AU - Schöppner, Volker AU - Hüllermeier, Eyke ID - 33090 JF - Welding in the World KW - Metals and Alloys KW - Mechanical Engineering KW - Mechanics of Materials SN - 0043-2288 TI - A comparison of heuristic, statistical, and machine learning methods for heated tool butt welding of two different materials ER - TY - CONF AB - In recent years, we observe an increasing amount of software with machine learning components being deployed. This poses the question of quality assurance for such components: how can we validate whether specified requirements are fulfilled by a machine learned software? Current testing and verification approaches either focus on a single requirement (e.g., fairness) or specialize on a single type of machine learning model (e.g., neural networks). In this paper, we propose property-driven testing of machine learning models. Our approach MLCheck encompasses (1) a language for property specification, and (2) a technique for systematic test case generation. The specification language is comparable to property-based testing languages. Test case generation employs advanced verification technology for a systematic, property dependent construction of test suites, without additional user supplied generator functions. We evaluate MLCheck using requirements and data sets from three different application areas (software discrimination, learning on knowledge graphs and security). Our evaluation shows that despite its generality MLCheck can even outperform specialised testing approaches while having a comparable runtime AU - Sharma, Arnab AU - Demir, Caglar AU - Ngonga Ngomo, Axel-Cyrille AU - Wehrheim, Heike ID - 28350 T2 - Proceedings of the 20th IEEE International Conference on Machine Learning and Applications (ICMLA) TI - MLCHECK–Property-Driven Testing of Machine Learning Classifiers ER - TY - JOUR AB - Automated machine learning (AutoML) supports the algorithmic construction and data-specific customization of machine learning pipelines, including the selection, combination, and parametrization of machine learning algorithms as main constituents. Generally speaking, AutoML approaches comprise two major components: a search space model and an optimizer for traversing the space. Recent approaches have shown impressive results in the realm of supervised learning, most notably (single-label) classification (SLC). Moreover, first attempts at extending these approaches towards multi-label classification (MLC) have been made. While the space of candidate pipelines is already huge in SLC, the complexity of the search space is raised to an even higher power in MLC. One may wonder, therefore, whether and to what extent optimizers established for SLC can scale to this increased complexity, and how they compare to each other. This paper makes the following contributions: First, we survey existing approaches to AutoML for MLC. Second, we augment these approaches with optimizers not previously tried for MLC. Third, we propose a benchmarking framework that supports a fair and systematic comparison. Fourth, we conduct an extensive experimental study, evaluating the methods on a suite of MLC problems. We find a grammar-based best-first search to compare favorably to other optimizers. AU - Wever, Marcel Dominik AU - Tornede, Alexander AU - Mohr, Felix AU - Hüllermeier, Eyke ID - 21004 JF - IEEE Transactions on Pattern Analysis and Machine Intelligence KW - Automated Machine Learning KW - Multi Label Classification KW - Hierarchical Planning KW - Bayesian Optimization SN - 0162-8828 TI - AutoML for Multi-Label Classification: Overview and Empirical Evaluation ER - TY - JOUR AB - Automated Machine Learning (AutoML) seeks to automatically find so-called machine learning pipelines that maximize the prediction performance when being used to train a model on a given dataset. One of the main and yet open challenges in AutoML is an effective use of computational resources: An AutoML process involves the evaluation of many candidate pipelines, which are costly but often ineffective because they are canceled due to a timeout. In this paper, we present an approach to predict the runtime of two-step machine learning pipelines with up to one pre-processor, which can be used to anticipate whether or not a pipeline will time out. Separate runtime models are trained offline for each algorithm that may be used in a pipeline, and an overall prediction is derived from these models. We empirically show that the approach increases successful evaluations made by an AutoML tool while preserving or even improving on the previously best solutions. AU - Mohr, Felix AU - Wever, Marcel Dominik AU - Tornede, Alexander AU - Hüllermeier, Eyke ID - 21092 JF - IEEE Transactions on Pattern Analysis and Machine Intelligence TI - Predicting Machine Learning Pipeline Runtimes in the Context of Automated Machine Learning ER - TY - CONF AU - Tornede, Tanja AU - Tornede, Alexander AU - Wever, Marcel Dominik AU - Hüllermeier, Eyke ID - 21570 T2 - Proceedings of the Genetic and Evolutionary Computation Conference TI - Coevolution of Remaining Useful Lifetime Estimation Pipelines for Automated Predictive Maintenance ER - TY - CONF AU - Hüllermeier, Eyke AU - Mohr, Felix AU - Tornede, Alexander AU - Wever, Marcel Dominik ID - 22913 TI - Automated Machine Learning, Bounded Rationality, and Rational Metareasoning ER - TY - GEN AB - Automated machine learning (AutoML) strives for the automatic configuration of machine learning algorithms and their composition into an overall (software) solution - a machine learning pipeline - tailored to the learning task (dataset) at hand. Over the last decade, AutoML has developed into an independent research field with hundreds of contributions. While AutoML offers many prospects, it is also known to be quite resource-intensive, which is one of its major points of criticism. The primary cause for a high resource consumption is that many approaches rely on the (costly) evaluation of many machine learning pipelines while searching for good candidates. This problem is amplified in the context of research on AutoML methods, due to large scale experiments conducted with many datasets and approaches, each of them being run with several repetitions to rule out random effects. In the spirit of recent work on Green AI, this paper is written in an attempt to raise the awareness of AutoML researchers for the problem and to elaborate on possible remedies. To this end, we identify four categories of actions the community may take towards more sustainable research on AutoML, i.e. Green AutoML: design of AutoML systems, benchmarking, transparency and research incentives. AU - Tornede, Tanja AU - Tornede, Alexander AU - Hanselle, Jonas Manuel AU - Wever, Marcel Dominik AU - Mohr, Felix AU - Hüllermeier, Eyke ID - 30866 T2 - arXiv:2111.05850 TI - Towards Green Automated Machine Learning: Status Quo and Future Directions ER - TY - THES AU - Wever, Marcel Dominik ID - 27284 TI - Automated Machine Learning for Multi-Label Classification ER - TY - CONF AU - Hanselle, Jonas Manuel AU - Tornede, Alexander AU - Wever, Marcel Dominik AU - Hüllermeier, Eyke ID - 21198 TI - Algorithm Selection as Superset Learning: Constructing Algorithm Selectors from Imprecise Performance Data ER - TY - CONF AU - Tornede, Alexander AU - Wever, Marcel Dominik AU - Hüllermeier, Eyke ID - 17407 T2 - Discovery Science TI - Extreme Algorithm Selection with Dyadic Feature Representation ER - TY - CONF AU - Hanselle, Jonas Manuel AU - Tornede, Alexander AU - Wever, Marcel Dominik AU - Hüllermeier, Eyke ID - 17408 T2 - KI 2020: Advances in Artificial Intelligence TI - Hybrid Ranking and Regression for Algorithm Selection ER - TY - CONF AU - Tornede, Tanja AU - Tornede, Alexander AU - Wever, Marcel Dominik AU - Mohr, Felix AU - Hüllermeier, Eyke ID - 17424 T2 - Proceedings of the ECMLPKDD 2020 TI - AutoML for Predictive Maintenance: One Tool to RUL Them All ER - TY - CONF AU - Tornede, Alexander AU - Wever, Marcel Dominik AU - Hüllermeier, Eyke ID - 20306 T2 - Workshop MetaLearn 2020 @ NeurIPS 2020 TI - Towards Meta-Algorithm Selection ER - TY - CONF AB - Algorithm selection (AS) deals with the automatic selection of an algorithm from a fixed set of candidate algorithms most suitable for a specific instance of an algorithmic problem class, where "suitability" often refers to an algorithm's runtime. Due to possibly extremely long runtimes of candidate algorithms, training data for algorithm selection models is usually generated under time constraints in the sense that not all algorithms are run to completion on all instances. Thus, training data usually comprises censored information, as the true runtime of algorithms timed out remains unknown. However, many standard AS approaches are not able to handle such information in a proper way. On the other side, survival analysis (SA) naturally supports censored data and offers appropriate ways to use such data for learning distributional models of algorithm runtime, as we demonstrate in this work. We leverage such models as a basis of a sophisticated decision-theoretic approach to algorithm selection, which we dub Run2Survive. Moreover, taking advantage of a framework of this kind, we advocate a risk-averse approach to algorithm selection, in which the avoidance of a timeout is given high priority. In an extensive experimental study with the standard benchmark ASlib, our approach is shown to be highly competitive and in many cases even superior to state-of-the-art AS approaches. AU - Tornede, Alexander AU - Wever, Marcel Dominik AU - Werner, Stefan AU - Mohr, Felix AU - Hüllermeier, Eyke ID - 18276 T2 - ACML 2020 TI - Run2Survive: A Decision-theoretic Approach to Algorithm Selection based on Survival Analysis ER - TY - CONF AB - In multi-label classification (MLC), each instance is associated with a set of class labels, in contrast to standard classification where an instance is assigned a single label. Binary relevance (BR) learning, which reduces a multi-label to a set of binary classification problems, one per label, is arguably the most straight-forward approach to MLC. In spite of its simplicity, BR proved to be competitive to more sophisticated MLC methods, and still achieves state-of-the-art performance for many loss functions. Somewhat surprisingly, the optimal choice of the base learner for tackling the binary classification problems has received very little attention so far. Taking advantage of the label independence assumption inherent to BR, we propose a label-wise base learner selection method optimizing label-wise macro averaged performance measures. In an extensive experimental evaluation, we find that or approach, called LiBRe, can significantly improve generalization performance. AU - Wever, Marcel Dominik AU - Tornede, Alexander AU - Mohr, Felix AU - Hüllermeier, Eyke ID - 15629 TI - LiBRe: Label-Wise Selection of Base Learners in Binary Relevance for Multi-Label Classification ER - TY - JOUR AB - In software engineering, the imprecise requirements of a user are transformed to a formal requirements specification during the requirements elicitation process. This process is usually guided by requirements engineers interviewing the user. We want to partially automate this first step of the software engineering process in order to enable users to specify a desired software system on their own. With our approach, users are only asked to provide exemplary behavioral descriptions. The problem of synthesizing a requirements specification from examples can partially be reduced to the problem of grammatical inference, to which we apply an active coevolutionary learning approach. However, this approach would usually require many feedback queries to be sent to the user. In this work, we extend and generalize our active learning approach to receive knowledge from multiple oracles, also known as proactive learning. The ‘user oracle’ represents input received from the user and the ‘knowledge oracle’ represents available, formalized domain knowledge. We call our two-oracle approach the ‘first apply knowledge then query’ (FAKT/Q) algorithm. We compare FAKT/Q to the active learning approach and provide an extensive benchmark evaluation. As result we find that the number of required user queries is reduced and the inference process is sped up significantly. Finally, with so-called On-The-Fly Markets, we present a motivation and an application of our approach where such knowledge is available. AU - Wever, Marcel Dominik AU - van Rooijen, Lorijn AU - Hamann, Heiko ID - 15025 IS - 2 JF - Evolutionary Computation TI - Multi-Oracle Coevolutionary Learning of Requirements Specifications from Examples in On-The-Fly Markets VL - 28 ER - TY - THES AU - Moussalem, Diego Campos ID - 16935 TI - Knowledge Graphs for Multilingual Language Translation and Generation ER - TY - JOUR AU - Karl, Holger AU - Kundisch, Dennis AU - Meyer auf der Heide, Friedhelm AU - Wehrheim, Heike ID - 13770 IS - 6 JF - Business & Information Systems Engineering TI - A Case for a New IT Ecosystem: On-The-Fly Computing VL - 62 ER - TY - GEN AU - Wever, Marcel Dominik AU - Mohr, Felix AU - Hüllermeier, Eyke AU - Hetzer, Alexander ID - 8868 TI - Towards Automated Machine Learning for Multi-Label Classification ER - TY - CONF AU - Melnikov, Vitaly AU - Hüllermeier, Eyke ID - 15007 T2 - Proceedings ACML, Asian Conference on Machine Learning (Proceedings of Machine Learning Research, 101) TI - Learning to Aggregate: Tackling the Aggregation/Disaggregation Problem for OWA ER - TY - CONF AU - Tornede, Alexander AU - Wever, Marcel Dominik AU - Hüllermeier, Eyke ED - Hoffmann, Frank ED - Hüllermeier, Eyke ED - Mikut, Ralf ID - 15011 SN - 978-3-7315-0979-0 T2 - Proceedings - 29. Workshop Computational Intelligence, Dortmund, 28. - 29. November 2019 TI - Algorithm Selection as Recommendation: From Collaborative Filtering to Dyad Ranking ER - TY - GEN AU - Mohr, Felix AU - Wever, Marcel Dominik AU - Tornede, Alexander AU - Hüllermeier, Eyke ID - 13132 T2 - INFORMATIK 2019: 50 Jahre Gesellschaft für Informatik – Informatik für Gesellschaft TI - From Automated to On-The-Fly Machine Learning ER - TY - CONF AB - Existing tools for automated machine learning, such as Auto-WEKA, TPOT, auto-sklearn, and more recently ML-Plan, have shown impressive results for the tasks of single-label classification and regression. Yet, there is only little work on other types of machine learning problems so far. In particular, there is almost no work on automating the engineering of machine learning solutions for multi-label classification (MLC). We show how the scope of ML-Plan, an AutoML-tool for multi-class classification, can be extended towards MLC using MEKA, which is a multi-label extension of the well-known Java library WEKA. The resulting approach recursively refines MEKA's multi-label classifiers, nesting other multi-label classifiers for meta algorithms and single-label classifiers provided by WEKA as base learners. In our evaluation, we find that the proposed approach yields strong results and performs significantly better than a set of baselines we compare with. AU - Wever, Marcel Dominik AU - Mohr, Felix AU - Tornede, Alexander AU - Hüllermeier, Eyke ID - 10232 TI - Automating Multi-Label Classification Extending ML-Plan ER - TY - CONF AU - Mohr, Felix AU - Wever, Marcel Dominik AU - Hüllermeier, Eyke AU - Faez, Amin ID - 2479 T2 - SCC TI - (WIP) Towards the Automated Composition of Machine Learning Services ER - TY - CONF AU - Mohr, Felix AU - Lettmann, Theodor AU - Hüllermeier, Eyke AU - Wever, Marcel Dominik ID - 2857 T2 - Proceedings of the 1st ICAPS Workshop on Hierarchical Planning TI - Programmatic Task Network Planning ER - TY - CONF AU - Mohr, Felix AU - Wever, Marcel Dominik AU - Hüllermeier, Eyke ID - 2471 T2 - SCC TI - On-The-Fly Service Construction with Prototypes ER - TY - JOUR AB - Automated machine learning (AutoML) seeks to automatically select, compose, and parametrize machine learning algorithms, so as to achieve optimal performance on a given task (dataset). Although current approaches to AutoML have already produced impressive results, the field is still far from mature, and new techniques are still being developed. In this paper, we present ML-Plan, a new approach to AutoML based on hierarchical planning. To highlight the potential of this approach, we compare ML-Plan to the state-of-the-art frameworks Auto-WEKA, auto-sklearn, and TPOT. In an extensive series of experiments, we show that ML-Plan is highly competitive and often outperforms existing approaches. AU - Mohr, Felix AU - Wever, Marcel Dominik AU - Hüllermeier, Eyke ID - 3510 JF - Machine Learning KW - AutoML KW - Hierarchical Planning KW - HTN planning KW - ML-Plan SN - 0885-6125 TI - ML-Plan: Automated Machine Learning via Hierarchical Planning ER - TY - CONF AU - Mohr, Felix AU - Wever, Marcel Dominik AU - Hüllermeier, Eyke ID - 3552 T2 - Proceedings of the Symposium on Intelligent Data Analysis TI - Reduction Stumps for Multi-Class Classification ER - TY - CONF AB - In automated machine learning (AutoML), the process of engineering machine learning applications with respect to a specific problem is (partially) automated. Various AutoML tools have already been introduced to provide out-of-the-box machine learning functionality. More specifically, by selecting machine learning algorithms and optimizing their hyperparameters, these tools produce a machine learning pipeline tailored to the problem at hand. Except for TPOT, all of these tools restrict the maximum number of processing steps of such a pipeline. However, as TPOT follows an evolutionary approach, it suffers from performance issues when dealing with larger datasets. In this paper, we present an alternative approach leveraging a hierarchical planning to configure machine learning pipelines that are unlimited in length. We evaluate our approach and find its performance to be competitive with other AutoML tools, including TPOT. AU - Wever, Marcel Dominik AU - Mohr, Felix AU - Hüllermeier, Eyke ID - 3852 KW - automated machine learning KW - complex pipelines KW - hierarchical planning T2 - ICML 2018 AutoML Workshop TI - ML-Plan for Unlimited-Length Machine Learning Pipelines ER - TY - CONF AB - In multinomial classification, reduction techniques are commonly used to decompose the original learning problem into several simpler problems. For example, by recursively bisecting the original set of classes, so-called nested dichotomies define a set of binary classification problems that are organized in the structure of a binary tree. In contrast to the existing one-shot heuristics for constructing nested dichotomies and motivated by recent work on algorithm configuration, we propose a genetic algorithm for optimizing the structure of such dichotomies. A key component of this approach is the proposed genetic representation that facilitates the application of standard genetic operators, while still supporting the exchange of partial solutions under recombination. We evaluate the approach in an extensive experimental study, showing that it yields classifiers with superior generalization performance. AU - Wever, Marcel Dominik AU - Mohr, Felix AU - Hüllermeier, Eyke ID - 2109 KW - Classification KW - Hierarchical Decomposition KW - Indirect Encoding T2 - Proceedings of the Genetic and Evolutionary Computation Conference, GECCO 2018, Kyoto, Japan, July 15-19, 2018 TI - Ensembles of Evolved Nested Dichotomies for Classification ER - TY - GEN AU - Wever, Marcel Dominik AU - Mohr, Felix AU - Hüllermeier, Eyke ID - 17713 TI - Automated Multi-Label Classification based on ML-Plan ER - TY - GEN AU - Mohr, Felix AU - Wever, Marcel Dominik AU - Hüllermeier, Eyke ID - 17714 TI - Automated machine learning service composition ER - TY - GEN AU - Graf, Helena ID - 5693 TI - Ranking of Classification Algorithms in AutoML ER - TY - GEN AU - Scheibl, Manuel ID - 5936 TI - Learning about learning curves from dataset properties ER - TY - CHAP AU - Schäfer, Dirk AU - Hüllermeier, Eyke ID - 6423 SN - 0302-9743 T2 - Discovery Science TI - Preference-Based Reinforcement Learning Using Dyad Ranking ER - TY - CONF AB - Whenever customers have to decide between different instances of the same product, they are interested in buying the best product. In contrast, companies are interested in reducing the construction effort (and usually as a consequence thereof, the quality) to gain profit. The described setting is widely known as opposed preferences in quality of the product and also applies to the context of service-oriented computing. In general, service-oriented computing emphasizes the construction of large software systems out of existing services, where services are small and self-contained pieces of software that adhere to a specified interface. Several implementations of the same interface are considered as several instances of the same service. Thereby, customers are interested in buying the best service implementation for their service composition wrt. to metrics, such as costs, energy, memory consumption, or execution time. One way to ensure the service quality is to employ certificates, which can come in different kinds: Technical certificates proving correctness can be automatically constructed by the service provider and again be automatically checked by the user. Digital certificates allow proof of the integrity of a product. Other certificates might be rolled out if service providers follow a good software construction principle, which is checked in annual audits. Whereas all of these certificates are handled differently in service markets, what they have in common is that they influence the buying decisions of customers. In this paper, we review state-of-the-art developments in certification with respect to service-oriented computing. We not only discuss how certificates are constructed and handled in service-oriented computing but also review the effects of certificates on the market from an economic perspective. AU - Jakobs, Marie-Christine AU - Krämer, Julia AU - van Straaten, Dirk AU - Lettmann, Theodor ED - Marcelo De Barros, Janusz Klink,Tadeus Uhl, Thomas Prinz ID - 115 T2 - The Ninth International Conferences on Advanced Service Computing (SERVICE COMPUTATION) TI - Certification Matters for Service Markets ER - TY - GEN AU - Schnitker, Nino Noel ID - 5694 TI - Genetischer Algorithmus zur Erstellung von Ensembles von Nested Dichotomies ER - TY - GEN AU - Hetzer, Alexander AU - Tornede, Tanja ID - 5724 TI - Solving the Container Pre-Marshalling Problem using Reinforcement Learning and Structured Output Prediction ER - TY - CONF AB - Today, software verification tools have reached the maturity to be used for large scale programs. Different tools perform differently well on varying code. A software developer is hence faced with the problem of choosing a tool appropriate for her program at hand. A ranking of tools on programs could facilitate the choice. Such rankings can, however, so far only be obtained by running all considered tools on the program.In this paper, we present a machine learning approach to predicting rankings of tools on programs. The method builds upon so-called label ranking algorithms, which we complement with appropriate kernels providing a similarity measure for programs. Our kernels employ a graph representation for software source code that mixes elements of control flow and program dependence graphs with abstract syntax trees. Using data sets from the software verification competition SV-COMP, we demonstrate our rank prediction technique to generalize well and achieve a rather high predictive accuracy (rank correlation > 0.6). AU - Czech, Mike AU - Hüllermeier, Eyke AU - Jakobs, Marie-Christine AU - Wehrheim, Heike ID - 71 T2 - Proceedings of the 3rd International Workshop on Software Analytics TI - Predicting Rankings of Software Verification Tools ER - TY - CONF AB - These days, there is a strong rise in the needs for machine learning applications, requiring an automation of machine learning engineering which is referred to as AutoML. In AutoML the selection, composition and parametrization of machine learning algorithms is automated and tailored to a specific problem, resulting in a machine learning pipeline. Current approaches reduce the AutoML problem to optimization of hyperparameters. Based on recursive task networks, in this paper we present one approach from the field of automated planning and one evolutionary optimization approach. Instead of simply parametrizing a given pipeline, this allows for structure optimization of machine learning pipelines, as well. We evaluate the two approaches in an extensive evaluation, finding both approaches to have their strengths in different areas. Moreover, the two approaches outperform the state-of-the-art tool Auto-WEKA in many settings. AU - Wever, Marcel Dominik AU - Mohr, Felix AU - Hüllermeier, Eyke ID - 1180 T2 - 27th Workshop Computational Intelligence TI - Automatic Machine Learning: Hierachical Planning Versus Evolutionary Optimization ER - TY - JOUR AB - Today, software components are provided by global markets in the form of services. In order to optimally satisfy service requesters and service providers, adequate techniques for automatic service matching are needed. However, a requester’s requirements may be vague and the information available about a provided service may be incomplete. As a consequence, fuzziness is induced into the matching procedure. The contribution of this paper is the development of a systematic matching procedure that leverages concepts and techniques from fuzzy logic and possibility theory based on our formal distinction between different sources and types of fuzziness in the context of service matching. In contrast to existing methods, our approach is able to deal with imprecision and incompleteness in service specifications and to inform users about the extent of induced fuzziness in order to improve the user’s decision-making. We demonstrate our approach on the example of specifications for service reputation based on ratings given by previous users. Our evaluation based on real service ratings shows the utility and applicability of our approach. AU - Platenius, Marie Christin AU - Shaker, Ammar AU - Becker, Matthias AU - Hüllermeier, Eyke AU - Schäfer, Wilhelm ID - 190 IS - 8 JF - IEEE Transactions on Software Engineering (TSE), presented at ICSE 2017 TI - Imprecise Matching of Requirements Specifications for Software Services using Fuzzy Logic ER - TY - CONF AB - Image Processing is fundamental for any camera-based vision system. In order to automate the prototyping process of image processing solutions to some extend, we propose a holistic, adaptive approach that comprises concepts for specification, composition, recommendation, execution, and rating of image processing functionality. The fundamental idea is to realize image processing applications according to Service-oriented Computing design principles. That is, distinct image processing functionality is encapsulated in terms of stateless services. Services are then used as building blocks for more complex image processing functionality. To automatically compose complex image processing functionality, our proposed approach incorporates a flexible, Artificial Intelligence planning-based forward search algorithm. Decision-making between alternative composition steps is supported by a learning recommendation system, which keeps track of valid composition steps by automatically constructing a composition grammar. In addition, it adapts to solutions of high quality by means of feedback-based Reinforcement Learning techniques. A concrete use case serves as proof of concept and demonstrates the feasibility of our holistic, adaptive approach. AU - Jungmann, Alexander AU - Kleinjohann, Bernd ID - 225 T2 - Proceedings of the 21st IEEE International Conference on Emerging Technologies and Factory Automation (ETFA) TI - A Holistic and Adaptive Approach for Automated Prototyping of Image Processing Functionality ER - TY - CONF AB - In the Image Processing domain, automated generation of complex Image Processing functionality is highly desirable; e.g., for rapid prototyping. Service composition techniques, in turn, facilitate automated generation of complex functionality based on building blocks in terms of services. For that reason, we aim for transferring the Service Composition paradigm into the Image Processing domain. In this paper, we present our symbolic composition approach that enables us to automatically generate Image Processing applications. Functionality of Image Processing services is described by means of a variant of first-order logic, which grounds on domain knowledge operationalized in terms of ontologies. A Petri-net formalism serves as basis for modeling data-flow of services and composed services. A planning-based composition algorithm automatically composes complex data-flow for a required functionality. A brief evaluation serves as proof of concept. AU - Jungmann, Alexander AU - Kleinjohann, Bernd ID - 218 T2 - Proceedings of the 13th IEEE International Conference on Services Computing (SCC) TI - Automatic Composition of Service-based Image Processing Applications ER - TY - THES AU - Jungmann, Alexander ID - 140 TI - Towards On-The-Fly Image Processing ER - TY - THES AU - Mohr, Felix ID - 141 TI - Towards Automated Service Composition Under Quality Constraints ER - TY - CONF AB - The Collaborative Research Centre "On-The-Fly Computing" works on foundations and principles for the vision of the Future Internet. It proposes the paradigm of On-The-Fly Computing, which tackles emerging worldwide service markets. In these markets, service providers trade software, platform, and infrastructure as a service. Service requesters state requirements on services. To satisfy these requirements, the new role of brokers, who are (human) actors building service compositions on the fly, is introduced. Brokers have to specify service compositions formally and comprehensively using a domain-specific language (DSL), and to use service matching for the discovery of the constituent services available in the market. The broker's choice of the DSL and matching approaches influences her success of building compositions as distinctive properties of different service markets play a significant role. In this paper, we propose a new approach of engineering a situation-specific DSL by customizing a comprehensive, modular DSL and its matching for given service market properties. This enables the broker to create market-specific composition specifications and to perform market-specific service matching. As a result, the broker builds service compositions satisfying the requester's requirements more accurately. We evaluated the presented concepts using case studies in service markets for tourism and university management. AU - Arifulina, Svetlana AU - Platenius, Marie Christin AU - Mohr, Felix AU - Engels, Gregor AU - Schäfer, Wilhelm ID - 280 T2 - Proceedings of the IEEE 11th World Congress on Services (SERVICES), Visionary Track: Service Composition for the Future Internet TI - Market-Specific Service Compositions: Specification and Matching ER - TY - CONF AB - In cloud computing, software architects develop systems for virtually unlimited resources that cloud providers account on a pay-per-use basis. Elasticity management systems provision these resources autonomously to deal with changing workload. Such changing workloads call for new objective metrics allowing architects to quantify quality properties like scalability, elasticity, and efficiency, e.g., for requirements/SLO engineering and software design analysis. In literature, initial metrics for these properties have been proposed. However, current metrics lack a systematic derivation and assume knowledge of implementation details like resource handling. Therefore, these metrics are inapplicable where such knowledge is unavailable.To cope with these lacks, this short paper derives metrics for scalability, elasticity, and efficiency properties of cloud computing systems using the goal question metric (GQM) method. Our derivation uses a running example that outlines characteristics of cloud computing systems. Eventually, this example allows us to set up a systematic GQM plan and to derive an initial set of six new metrics. We particularly show that our GQM plan allows to classify existing metrics. AU - Becker, Matthias AU - Lehrig, Sebastian AU - Becker, Steffen ID - 245 T2 - Proceedings of the 6th ACM/SPEC International Conference on Performance Engineering TI - Systematically Deriving Quality Metrics for Cloud Computing Systems ER - TY - JOUR AB - On-the-fly composition of service-based software solutions is still a challenging task. Even more challenges emerge when facing automatic service composition in markets of composed services for end users. In this paper, we focus on the functional discrepancy between “what a user wants” specified in terms of a request and “what a user gets” when executing a composed service. To meet the challenge of functional discrepancy, we propose the combination of existing symbolic composition approaches with machine learning techniques. We developed a learning recommendation system that expands the capabilities of existing composition algorithms to facilitate adaptivity and consequently reduces functional discrepancy. As a representative of symbolic techniques, an Artificial Intelligence planning based approach produces solutions that are correct with respect to formal specifications. Our learning recommendation system supports the symbolic approach in decision-making. Reinforcement Learning techniques enable the recommendation system to adjust its recommendation strategy over time based on user ratings. We implemented the proposed functionality in terms of a prototypical composition framework. Preliminary results from experiments conducted in the image processing domain illustrate the benefit of combining both complementary techniques. AU - Jungmann, Alexander AU - Mohr, Felix ID - 323 IS - 1 JF - Journal of Internet Services and Applications TI - An approach towards adaptive service composition in markets of composed services ER - TY - CONF AB - Services are self-contained software components that can beused platform independent and that aim at maximizing software reuse. Abasic concern in service oriented architectures is to measure the reusabilityof services. One of the most important qualities is the functionalreusability, which indicates how relevant the task is that a service solves.Current metrics for functional reusability of software, however, have verylittle explanatory power and do not accomplish this goal.This paper presents a new approach to estimate the functional reusabilityof services based on their relevance. To this end, it denes the degreeto which a service enables the execution of other services as its contri-bution. Based on the contribution, relevance of services is dened as anestimation for their functional reusability. Explanatory power is obtainedby normalizing relevance values with a reference service. The applicationof the metric to a service test set conrms its supposed capabilities. AU - Mohr, Felix ID - 324 T2 - Proceedings of the 14th International Conference on Software Reuse (ICSR) TI - A Metric for Functional Reusability of Services ER - TY - JOUR AB - In this paper we consider an extended variant of query learning where the hidden concept is embedded in some Boolean circuit. This additional processing layer modifies query arguments and answers by fixed transformation functions which are known to the learner. For this scenario, we provide a characterization of the solution space and an ordering on it. We give a compact representation of the minimal and maximal solutions as quantified Boolean formulas and we adapt the original algorithms for exact learning of specific classes of propositional formulas. AU - Bubeck, Uwe AU - Kleine Büning, Hans ID - 3343 JF - Artificial Intelligence KW - Query learning KW - Propositional logic SN - 0004-3702 TI - Learning Boolean Specifications ER - TY - CONF AB - In this paper, we introduce an approach for combining embedded systems with Service-oriented Computing techniques based on a concrete application scenario from the robotics domain. Our proposed Service-oriented Architecture allows for incorporating computational expensive functionality as services into a distributed computing environment. Furthermore, our framework facilitates a seamless integration of embedded systems such as robots as service providers into the computing environment. The entire communication is based on so-called recipes, which can be interpreted as autonomous messages that contain all necessary information for executing compositions of services. AU - Jungmann, Alexander AU - Jatzkowski, Jan AU - Kleinjohann, Bernd ID - 315 T2 - Proceedings of the 5th IFIP International Embedded Systems Symposium TI - Combining Service-oriented Computing with Embedded Systems - A Robotics Case Study ER - TY - CONF AB - Services are self-contained and platform independent software components that aim at maximizing software reuse. The automated composition of services to a target software artifact has been tackled with many AI techniques, but existing approaches make unreasonably strong assumptions such as a predefined data flow, are limited to tiny problem sizes, ignore non-functional properties, or assume offline service repositories. This paper presents an algorithm that automatically composes services without making such assumptions. We employ a backward search algorithm that starts from an empty composition and prepends service calls to already discovered candidates until a solution is found. Available services are determined during the search process. We implemented our algorithm, performed an experimental evaluation, and compared it to other approaches. AU - Mohr, Felix AU - Jungmann, Alexander AU - Kleine Büning, Hans ID - 319 T2 - Proceedings of the 12th IEEE International Conference on Services Computing (SCC) TI - Automated Online Service Composition ER - TY - CONF AB - Automated service composition aims at automatically generating software solutions based on services to provide more complex functionality. In this paper, we give an initial overview about why adaptivity becomes increasingly important when aiming for automated composition of service functionality in dynamic and freely accessible environments such as service markets. We systematically derive dependencies among crucial processes such as service composition and service execution in a holistic view. Furthermore, we briefly discuss the influences and effects of changes in the environment according to the derived dependencies, and derive possible future research directions. AU - Jungmann, Alexander ID - 272 T2 - Proceedings of the IEEE 11th World Congress on Services (SERVICES) TI - On Adaptivity for Automated Composition of Service Functionality ER - TY - CONF AB - Automatically composing service-based software solutions is a challenging task. Considering context information during this service composition process is even more challenging. In domains such as image processing, however, context-sensitivity is inherent and cannot be ignored when developing techniques for automatic service composition. Formal approaches tend to create ambiguous solutions, whenever the expressive power of the applied formalism is limited. For example, services may have the same formal specification, although their actual functionality depends on the concrete context. In order to satisfy individual user requests while providing data-dependent functionality, formal approaches have to be extended. We propose to incorporate Reinforcement Learning techniques and combine them with planning based composition approaches. While planning ensures formally correct solutions, learning enables the composition process to resolve ambiguity by implicitly considering context information. Preliminary results show that our combined approach adapts to a static context while still satisfying formally specified requirements. AU - Jungmann, Alexander AU - Kleinjohann, Bernd ID - 345 T2 - Proceedings of the 6th International Conference on Cloud Computing Technology and Science (CloudCom) TI - Towards Context-Sensitive Service Composition for Service-Oriented Image Processing ER - TY - CONF AB - One future goal of service-oriented computing is to realize global markets of composed services. On such markets, service providers offer services that can be flexibly combined with each other. However, most often, market participants are not able to individually estimate the quality of traded services in advance. As a consequence, even potentially profitable transactions between customers and providers might not take place. In the worst case, this can induce a market failure. To overcome this problem, we propose the incorporation of reputation information as an indicator for expected service quality. We address On-The-Fly Computing as a representative environment of markets of composed services. In this environment, customers provide feedback on transactions. We present a conceptual design of a reputation system which collects and processes user feedback, and provides it to participants in the market. Our contribution includes the identification of requirements for such a reputation system from a technical and an economic perspective. Based on these requirements, we propose a flexible solution that facilitates the incorporation of reputation information into markets of composed services while simultaneously preserving privacy of customers who provide feedback. The requirements we formulate in this paper have just been partially met in literature. An integrated approach, however, has not been addressed yet. AU - Brangewitz, Sonja AU - Jungmann, Alexander AU - Petrlic, Ronald AU - Platenius, Marie Christin ID - 346 T2 - Proceedings of the 6th International Conferences on Advanced Service Computing (SERVICE COMPUTATION) TI - Towards a Flexible and Privacy-Preserving Reputation System for Markets of Composed Services ER - TY - CONF AB - There are many technologies for the automation of processesthat deal with services; examples are service discovery and composition.Automation of these processes requires that the services are described semantically. However, semantically described services are currently not oronly rarely available, which limits the applicability of discovery and composition approaches. The systematic support for creating new semanticservices usable by automated technologies is an open problem.We tackle this problem with a template based approach: Domain independent templates are instantiated with domain specific services andboolean expressions. The obtained services have semantic descriptionswhose correctness directly follows from the correctness of the template.Besides the theory, we present experimental results for a service repository in which 85% of the services were generated automatically. AU - Mohr, Felix AU - Walther, Sven ID - 353 T2 - Proceedings of the 14th International Conference on Software Reuse (ICSR) TI - Template-based Generation of Semantic Services ER - TY - CONF AB - On-The-Fly (OTF) Computing constitutes an approach towards highly dynamic and individualized software markets. Based on service-oriented computing, OTF Computing is about realizing global markets of services that can be flexibly combined. We report on our current research activities, the security and privacy implications thereof, and our approaches to tackle the challenges. Furthermore, we discuss how the security and privacy challenges are addressed in research projects similar to OTF Computing. AU - Petrlic, Ronald AU - Jungmann, Alexander AU - Platenius, Marie Christin AU - Schäfer, Wilhelm AU - Sorge, Christoph ID - 366 T2 - Tagungsband der 4. Konferenz Software-Technologien und -Prozesse (STeP 2014) TI - Security and Privacy Challenges in On-The-Fly Computing ER - TY - CONF AB - Automatic service composition is still a challengingtask. It is even more challenging when dealing witha dynamic market of services for end users. New servicesmay enter the market while other services are completelyremoved. Furthermore, end users are typically no experts in thedomain in which they formulate a request. As a consequence,ambiguous user requests will inevitably emerge and have tobe taken into account. To meet these challenges, we proposea new approach that combines automatic service compositionwith adaptive service recommendation. A best first backwardsearch algorithm produces solutions that are functional correctwith respect to user requests. An adaptive recommendationsystem supports the search algorithm in decision-making.Reinforcement Learning techniques enable the system to adjustits recommendation strategy over time based on user ratings.The integrated approach is described on a conceptional leveland demonstrated by means of an illustrative example fromthe image processing domain. AU - Jungmann, Alexander AU - Mohr, Felix AU - Kleinjohann, Bernd ID - 447 T2 - Proceedings of the 10th World Congress on Services (SERVICES) TI - Combining Automatic Service Composition with Adaptive Service Recommendation for Dynamic Markets of Services ER - TY - GEN AU - Heldt, Waleri ID - 454 TI - Automated Service Composition: Adaption of the ASTRO Approach ER - TY - CONF AB - Automatically composing service-based software solutionsis still a challenging task. Functional as well as nonfunctionalproperties have to be considered in order to satisfyindividual user requests. Regarding non-functional properties,the composition process can be modeled as optimization problemand solved accordingly. Functional properties, in turn, can bedescribed by means of a formal specification language. Statespacebased planning approaches can then be applied to solvethe underlying composition problem. However, depending on theexpressiveness of the applied formalism and the completenessof the functional descriptions, formally equivalent services maystill differ with respect to their implemented functionality. As aconsequence, the most appropriate solution for a desired functionalitycan hardly be determined without considering additionalinformation. In this paper, we demonstrate how to overcome thislack of information by means of Reinforcement Learning. Inorder to resolve ambiguity, we expand state-space based servicecomposition by a recommendation mechanism that supportsdecision-making beyond formal specifications. The recommendationmechanism adjusts its recommendation strategy basedon feedback from previous composition runs. Image processingserves as case study. Experimental results show the benefit of ourproposed solution. AU - Jungmann, Alexander AU - Mohr, Felix AU - Kleinjohann, Bernd ID - 457 T2 - Proceedings of the 7th International Conference on Service Oriented Computing and Applications (SOCA) TI - Applying Reinforcement Learning for Resolving Ambiguity in Service Composition ER - TY - CONF AB - Automated programming aims at automatically assembling a new software artifact from existing software modules. Although automated programming was revitalized through automated software composition in the last decade, the problem cannot be considered solved. Automated software composition is widely accepted as being a planning task, but the problem is that it has very special properties that other planning problems do not have and that are commonly overseen. These properties usually imply that the composition problem cannot be solved with standard planning tools. This paper gives a brief and intuitive description of the planning problem that most approaches are based on. It points out special properties of this problem and explains why it is not adequate to solve the problem with classical planning tools as done by most existing approaches. AU - Mohr, Felix ID - 407 T2 - Proceedings of the 29th International Conference on Automated Software Engineering (ASE) TI - Issues of Automated Software Composition in AI Planning ER - TY - JOUR AB - One goal of service-oriented computing is to realize future markets of composed services. In such markets, service providers offer services that can be flexibly combined with each other. However, although crucial for decision-making, market participants are usually not able to individually estimate the quality of traded services in advance. To overcome this problem, we present a conceptual design for a reputation system that collects and processes user feedback on transactions, and provides this information as a signal for quality to participants in the market. Based on our proposed concept, we describe the incorporation of reputation information into distinct decision-making processes that are crucial in such service markets. In this context, we present a fuzzy service matching approach that takes reputation information into account. Furthermore, we introduce an adaptive service composition approach, and investigate the impact of exchanging immediate user feedback by reputation information. Last but not least, we describe the importance of reputation information for economic decisions of different market participants. The overall output of this paper is a comprehensive view on managing and exploiting reputation information in markets of composed services using the example of On-The-Fly Computing. AU - Jungmann, Alexander AU - Brangewitz, Sonja AU - Petrlic, Ronald AU - Platenius, Marie Christin ID - 410 IS - 3&4 JF - International Journal On Advances in Intelligent Systems (IntSys) TI - Incorporating Reputation Information into Decision-Making Processes in Markets of Composed Services VL - 7 ER - TY - GEN AU - Finkensiep, Christoph ID - 424 TI - Fast and Flexible Automatic Composition of Semantic Web Services ER - TY - CONF AB - In this paper, we evaluate the robustness of our color-based segmentation approach in combination with different color spaces, namely RGB, L*a*b*, HSV, and log-chromaticity (LCCS). For this purpose, we describe our deterministic segmentation algorithm including its gradually transformation of pixel-precise image data into a less error-prone and therefore more robust statistical representation in terms of moments. To investigate the robustness of a specific segmentation setting, we introduce our evaluation framework that directly works on the statistical representation. It is based on two different types of robustness measures, namely relative and absolute robustness. While relative robustness measures stability of segmentation results over time, absolute robustness measures stability regarding varying illumination by comparing results with ground truth data. The significance of these robustness measures is shown by evaluating our segmentation approach with different color spaces. For the evaluation process, an artificial scene was chosen as representative for application scenarios based on artificial landmarks. AU - Jungmann, Alexander AU - Jatzkowski, Jan AU - Kleinjohann, Bernd ID - 425 T2 - Proceedings of the 9th International Conference on Computer Vision Theory and Applications (VISAPP) TI - Evaluation of Color Spaces for Robust Image Segmentation ER - TY - CONF AB - Services are self-contained software components that can be used platform independent and that aim at maximizing software reuse. A basic concern in service oriented architectures is to measure the reusability of services. One of the most important qualities is the functional reusability, which indicates how relevant the task is that a service solves. Current metrics for functional reusability of software, however, either require source code analysis or have very little explanatory power. This paper gives a formally described vision statement for the estimation of functional reusability of services and sketches an exemplary reusability metric that is based on the service descriptions. AU - Mohr, Felix ID - 428 T2 - Proceedings of the 12th International Conference on Service Oriented Computing (ICSOC) TI - Estimating Functional Reusability of Services ER - TY - GEN AU - Bieshaar, Maarten ID - 482 TI - Statistisches Planen von Aktionen für autonome mobile Roboter in realen Umgebungen ER - TY - CONF AB - Software composition has been studied as a subject of state based planning for decades. Existing composition approaches that are efficient enough to be used in practice are limited to sequential arrangements of software components. This restriction dramatically reduces the number of composition problems that can be solved. However, there are many composition problems that could be solved by existing approaches if they had a possibility to combine components in very simple non-sequential ways. To this end, we present an approach that arranges not only basic components but also composite components. Composite components enhance the structure of the composition by conditional control flows. Through algorithms that are written by experts, composite components are automatically generated before the composition process starts. Therefore, our approach is not a substitute for existing composition algorithms but complements them with a preprocessing step. We verified the validity of our approach through implementation of the presented algorithms. AU - Mohr, Felix AU - Kleine Büning, Hans ID - 485 T2 - Proceedings of the 15th International Conference on Information Integration and Web-based Applications & Services (iiWAS) TI - Semi-Automated Software Composition Through Generated Components ER - TY - CONF AB - Automated service composition has been studied as a subject of state based planning for a decade. A great deal of service composition tasks can only be solved if concrete output values of the services are considered in the composition process. However, the fact that those values are not known before runtime leads to nondeterministic planning problems, which have proven to be notoriously difficult in practical automated service composition applications. Even though this problem is frequently recognized, it has still received remarkably few attention and remains unsolved.This paper shows how nondeterminism in automated service composition can be reduced. We introduce context rules as a means to derive semantic knowledge from output values of services. These rules enable us to replace nondeterministic composition operations by less nondeterministic or even completely deterministic ones. We show the validity of our solutions not only theoretically but also have evaluated them practically through implementation. AU - Mohr, Felix AU - Lettmann, Theodor AU - Kleine Büning, Hans ID - 495 T2 - Proceedings of the 6th International Conference on Service Oriented Computing and Applications (SOCA) TI - Reducing Nondeterminism in Automated Service Composition ER - TY - JOUR AB - The as a service paradigm reflects the fundamental idea of providing basic coherent functionality in terms of components that can be utilised on demand. These so-called services may also be interconnected in order to provide more complex functionality. Automation of this service composition process is indeed a formidable challenge. In our work, we are addressing this challenge by decomposing service composition into sequential decision making steps. Each step is supported by a recommendation mechanism. If composition requests recur over time and if evaluations of composition results are fed back, a proper recommendation strategy can evolve over time through learning from experience. In this paper, we describe our approach of modelling this service composition and recommendation process as Markov decision process and of solving it by means of reinforcement learning. A case study serves as proof of concept. AU - Jungmann, Alexander AU - Kleinjohann, Bernd AU - Kleinjohann, Elisabeth ID - 515 IS - 4 JF - International Journal of Business Process Integration and Management TI - Learning Service Recommendations ER - TY - CONF AB - The as a Service paradigm reflects the fundamental idea of providing basic coherent functionality in terms of components that can be utilized on demand. These so-called services may also be interconnected in order to provide more complex functionality. Automation of this service composition process is indeed a formidable challenge. In our work, we are addressing this challenge by decomposing service composition into sequential decision making steps. Each step is supported by a recommendation mechanism. If composition requests recur over time and if evaluations of composition results are fed back, a proper recommendation strategy can evolve over time through learning from experience. In this paper, we describe our general idea of modeling this service composition and recommendation process as Markov Decision Process and of solving it by means of Reinforcement Learning. A case study serves as proof of concept. AU - Jungmann, Alexander AU - Kleinjohann, Bernd ID - 516 T2 - Proceedings of the 10th IEEE International Conference on Services Computing (SCC) TI - Learning Recommendation System for Automated Service Composition ER - TY - GEN AU - Buse, Dominik ID - 530 TI - Entwurf kooperativer Verhaltensweisen heterogener Roboter ER - TY - GEN AU - Borkowski, Richard ID - 533 TI - Entwicklung eines Hybriden Planers zur verhaltensorientierten Selbstoptimierung ER - TY - CONF AB - A major goal of the On-The-Fly Computing project is the automated composition of individual services based on services that are available in dynamic markets. Dependent on the granularity of a market, different alternatives that satisfy the requested functional requirements may emerge. In order to select the best solution, services are usually selected with respect to their quality in terms of inherent non-functional properties. In this paper, we describe our idea of how to model this service selection process as a Markov Decision Process, which we in turn intend to solve by means of Reinforcement Learning techniques in order to control the underlying service composition process. In addition, some initial issues with respect to our approach are addressed. AU - Jungmann, Alexander AU - Kleinjohann, Bernd ID - 568 T2 - Proceedings of the 9th IEEE International Conference on Service Computing (SCC) TI - Towards the Application of Reinforcement Learning Techniques for Quality-Based Service Selection in Automated Service Composition ER - TY - CONF AB - The paradigm shift from purchasing monolithic software solutions to a dynamic composition of individual solutions entails many new possibilities yet great challenges, too. In order to satisfy user requirements, complex services have to be automatically composed of elementary services. Multiple possibilities of composing a complex service inevitably emerge. The problem of selecting the most appropriate services has to be solved by comparing the different service candidates with respect to their quality in terms of inherent non-functional properties while simultaneously taking the user requirements into account. We are aiming for an integrated service rating and ranking methodology in order to support the automation of the underlying decision-making process. The main contribution of this paper is a first decomposition of the quality-based service selection process, while emphasizing major issues and challenges, which we are addressing in the On-The-Fly Computing project. AU - Jungmann, Alexander AU - Kleinjohann, Bernd ID - 571 T2 - Proceedings of the 4th International Conferences on Advanced Service Computing (SERVICE COMPUTATION) TI - Towards an Integrated Service Rating and Ranking Methodology for Quality Based Service Selection in Automatic Service Composition ER - TY - CONF AB - In this paper, a color based feature extraction and classification approach for image processing in embedded systems in presented. The algorithms and data structures developed for this approach pay particular attention to reduce memory consumption and computation power of the entire image processing, since embedded systems usually impose strong restrictions regarding those resources. The feature extraction is realized in terms of an image segmentation algorithm. The criteria of homogeneity for merging pixels and regions is provided by the color classification mechanism, which incorporates appropriate methods for defining, representing and accessing subspaces in the working color space. By doing so, pixels and regions with color values that belong to the same color class can be merged. Furthermore, pixels with redundant color values that do not belong to any pre-defined color class can be completely discarded in order to minimize computational effort. Subsequently, the extracted regions are converted to a more convenient feature representation in terms of statistical moments up to and including second order. For evaluation, the whole image processing approach is applied to a mobile representative of embedded systems within the scope of a simple real-world scenario. AU - Jungmann, Alexander AU - Kleinjohann, Bernd AU - Kleinjohann, Elisabeth AU - Bieshaar, Maarten ID - 617 T2 - Proceedings of the Fourth International Conference on Resource Intensive Applications and Services (INTENSIVE) TI - Efficient Color-Based Image Segmentation and Feature Classification for Image Processing in Embedded Systems ER -