摘要 :
Recent trends like the (Industrial) Internet of Things and Industry 4.0 lead to highly integrated machines and thus to greater challenges in dealing with data, mostly with respect to its volume and velocity. It is impossible to co...
展开
Recent trends like the (Industrial) Internet of Things and Industry 4.0 lead to highly integrated machines and thus to greater challenges in dealing with data, mostly with respect to its volume and velocity. It is impossible to collect all data available, both at maximum breadth (number of values) and maximum depth (frequency and precision). The goal is to achieve an optimal trade-off between bandwidth utilization versus information transmitted. This requires optimized data collection strategies, which can extensively profit from involving the domain expert's knowledge about the process. In this paper, we build on our previously presented optimized data load methods, that leverage process-driven data collection. These enable data providers (ⅰ) to split their production process into phases, (ⅱ) for each phase to precisely define what data to collect and how and (ⅲ) to model transitions between phases via a data-driven method. This paper extends the previous approach in both breadth and depth and focuses especially on making its benefits, like the demonstrated 39% savings in bandwidth, to domain experts. We propose a novel, user-friendly assistant that enables domain experts to define, deploy and maintain a flexible data integration pipeline from the edge of production to the cloud.
收起
摘要 :
Recent trends like the (Industrial) Internet of Things and Industry 4.0 lead to highly integrated machines and thus to greater challenges in dealing with data, mostly with respect to its volume and velocity. It is impossible to co...
展开
Recent trends like the (Industrial) Internet of Things and Industry 4.0 lead to highly integrated machines and thus to greater challenges in dealing with data, mostly with respect to its volume and velocity. It is impossible to collect all data available, both at maximum breadth (number of values) and maximum depth (frequency and precision). The goal is to achieve an optimal trade-off between bandwidth utilization versus information transmitted. This requires optimized data collection strategies, which can extensively profit from involving the domain expert's knowledge about the process. In this paper, we build on our previously presented optimized data load methods, that leverage process-driven data collection. These enable data providers (ⅰ) to split their production process into phases, (ⅱ) for each phase to precisely define what data to collect and how and (ⅲ) to model transitions between phases via a data-driven method. This paper extends the previous approach in both breadth and depth and focuses especially on making its benefits, like the demonstrated 39% savings in bandwidth, to domain experts. We propose a novel, user-friendly assistant that enables domain experts to define, deploy and maintain a flexible data integration pipeline from the edge of production to the cloud.
收起
摘要 :
[Context & Motivation] Once research questions and initial theories have shaped, empirical research typically requires to select cases to study subsumed ideas. Issue trackers of todays open source systems (OSS) are a gold mine for empirical research, not least to study trace links among the included issue artifacts. [Question/problem] The huge amount of available OSS projects complicates the process of finding suitable cases to support the research goals. Further, simply picking a large number of projects on a random basis does not imply generalizability. Therefore the selection process should be carefully designed. [Principle ideas/results] In this paper we propose a method to choose OSS projects to study trace links found in issue tracking systems. Builds upon purposive sampling and cluster analysis, relevant project characteristics are identified whereas irrelevant information is filtered. Every step of the method is demonstrated on a live example. [Contributions] The proposed strategy selects an information-rich, representative and diverse sample of OSS to perform a traceability case study. Our work may be used as practical guide for other researchers to perform project selection tasks....
展开
[Context & Motivation] Once research questions and initial theories have shaped, empirical research typically requires to select cases to study subsumed ideas. Issue trackers of todays open source systems (OSS) are a gold mine for empirical research, not least to study trace links among the included issue artifacts. [Question/problem] The huge amount of available OSS projects complicates the process of finding suitable cases to support the research goals. Further, simply picking a large number of projects on a random basis does not imply generalizability. Therefore the selection process should be carefully designed. [Principle ideas/results] In this paper we propose a method to choose OSS projects to study trace links found in issue tracking systems. Builds upon purposive sampling and cluster analysis, relevant project characteristics are identified whereas irrelevant information is filtered. Every step of the method is demonstrated on a live example. [Contributions] The proposed strategy selects an information-rich, representative and diverse sample of OSS to perform a traceability case study. Our work may be used as practical guide for other researchers to perform project selection tasks.
收起
摘要 :
Approximate inference in large and densely connected graphical models is a challenging but highly relevant problem. Belief propagation, as a method for performing approximate inference in loopy graphs, has shown empirical success ...
展开
Approximate inference in large and densely connected graphical models is a challenging but highly relevant problem. Belief propagation, as a method for performing approximate inference in loopy graphs, has shown empirical success in many applications. However, convergence of belief propagation can only be guaranteed for simple graphs. Whether belief propagation converges depends strongly on the applied message update scheme, and specialized schemes can be highly beneficial. Yet, residual belief propagation is the only established method utilizing this fact to improve convergence properties. In experiments, we observe that residual belief propagation fails to converge if local oscillations occur and the same sequence of messages is repeatedly updated. To overcome this issue, we propose two novel message update schemes. In the first scheme we add noise to oscillating messages. In the second scheme we apply weight decay to gradually reduce the influence of these messages and consequently enforce convergence. Furthermore, in contrast to previous work, we consider the correctness of the obtained marginals and observe significant performance improvements when applying the proposed message update schemes to various Ising models with binary random variables.
收起
摘要 :
Approximate inference in large and densely connected graphical models is a challenging but highly relevant problem. Belief propagation, as a method for performing approximate inference in loopy graphs, has shown empirical success ...
展开
Approximate inference in large and densely connected graphical models is a challenging but highly relevant problem. Belief propagation, as a method for performing approximate inference in loopy graphs, has shown empirical success in many applications. However, convergence of belief propagation can only be guaranteed for simple graphs. Whether belief propagation converges depends strongly on the applied message update scheme, and specialized schemes can be highly beneficial. Yet, residual belief propagation is the only established method utilizing this fact to improve convergence properties. In experiments, we observe that residual belief propagation fails to converge if local oscillations occur and the same sequence of messages is repeatedly updated. To overcome this issue, we propose two novel message update schemes. In the first scheme we add noise to oscillating messages. In the second scheme we apply weight decay to gradually reduce the influence of these messages and consequently enforce convergence. Furthermore, in contrast to previous work, we consider the correctness of the obtained marginals and observe significant performance improvements when applying the proposed message update schemes to various Ising models with binary random variables.
收起
摘要 :
[Context & Motivation] Once research questions and initial theories have shaped, empirical research typically requires to select cases to study subsumed ideas. Issue trackers of todays open source systems (OSS) are a gold mine for empirical research, not least to study trace links among the included issue artifacts. [Question / problem] The huge amount of available OSS projects complicates the process of finding suitable cases to support the research goals. Further, simply picking a large number of projects on a random basis does not imply gener-alizability. Therefore the selection process should be carefully designed. [Principle ideas / results] In this paper we propose a method to choose OSS projects to study trace links found in issue tracking systems. Builds upon purposive sampling and cluster analysis, relevant project characteristics are identified whereas irrelevant information is filtered. Every step of the method is demonstrated on a live example. [Contributions] The proposed strategy selects an information-rich, representative and diverse sample of OSS to perform a traceability case study. Our work may be used as practical guide for other researchers to perform project selection tasks....
展开
[Context & Motivation] Once research questions and initial theories have shaped, empirical research typically requires to select cases to study subsumed ideas. Issue trackers of todays open source systems (OSS) are a gold mine for empirical research, not least to study trace links among the included issue artifacts. [Question / problem] The huge amount of available OSS projects complicates the process of finding suitable cases to support the research goals. Further, simply picking a large number of projects on a random basis does not imply gener-alizability. Therefore the selection process should be carefully designed. [Principle ideas / results] In this paper we propose a method to choose OSS projects to study trace links found in issue tracking systems. Builds upon purposive sampling and cluster analysis, relevant project characteristics are identified whereas irrelevant information is filtered. Every step of the method is demonstrated on a live example. [Contributions] The proposed strategy selects an information-rich, representative and diverse sample of OSS to perform a traceability case study. Our work may be used as practical guide for other researchers to perform project selection tasks.
收起
摘要 :
Traceability, a classic requirements engineering topic, is increasingly used in the context of model-based engineering. However, researchers and practitioners lack a concise terminology to discuss aspects of requirements traceabil...
展开
Traceability, a classic requirements engineering topic, is increasingly used in the context of model-based engineering. However, researchers and practitioners lack a concise terminology to discuss aspects of requirements traceability in situations in which engineers heavily rely on models and model-based engineering. While others have previously surveyed the domain, no one has so far provided a clear, unambiguous set of terms that can be used to discuss traceability in such a context. We therefore set out to cut a path through the jungle of terminology for model-based traceability, ground it in established terminology from requirements engineering, and derive an unambiguous set of relevant terms. We also map the terminology used in existing primary and secondary studies to our taxonomy to show differences and commonalities. The contribution of this paper is thus a terminology for model-based traceability that allows requirements engineers and engineers working with models to unambiguously discuss their joint traceability efforts.
收起
摘要 :
Traceability, a classic requirements engineering topic, is increasingly used in the context of model-based engineering. However, researchers and practitioners lack a concise terminology to discuss aspects of requirements traceabil...
展开
Traceability, a classic requirements engineering topic, is increasingly used in the context of model-based engineering. However, researchers and practitioners lack a concise terminology to discuss aspects of requirements traceability in situations in which engineers heavily rely on models and model-based engineering. While others have previously surveyed the domain, no one has so far provided a clear, unambiguous set of terms that can be used to discuss traceability in such a context. We therefore set out to cut a path through the jungle of terminology for model-based traceability, ground it in established terminology from requirements engineering, and derive an unambiguous set of relevant terms. We also map the terminology used in existing primary and secondary studies to our taxonomy to show differences and commonalities. The contribution of this paper is thus a terminology for model-based traceability that allows requirements engineers and engineers working with models to unambiguously discuss their joint traceability efforts.
收起
摘要 :
A large body of work has shown that ultra-wideband (UWB) technology enables accurate indoor localization and tracking thanks to its high time-domain resolution. Existing systems, however, are typically designed to localize only a ...
展开
A large body of work has shown that ultra-wideband (UWB) technology enables accurate indoor localization and tracking thanks to its high time-domain resolution. Existing systems, however, are typically designed to localize only a limited number of tags, and involve the exchange of several messages following a given schedule. As a result, the scalability of current solutions in terms of tag density is limited, as well as their efficiency and responsiveness. In this paper, we present SnapLoc, a UWB-based indoor localization system that allows an unlimited number of tags to self-localize at a theoretical upper bound of 2.3 kHz. In SnapLoc, a tag obtains the responses from multiple anchors simultaneously. Based on these signals, the tag derives the time difference of arrival between anchors and estimates its position. Therefore, SnapLoc does not require tags to actively transmit packets, but to receive only a single message. This allows tags to passively localize themselves and ensures that the performance of SnapLoc does not degrade with high node densities. Moreover, due to the (quasi-)simultaneous responses, a tight clock synchronization between anchors is not needed. We have implemented SnapLoc on a low-cost platform based on the De-cawave DW1000 radio and solved limitations in the transceiver's timestamp resolution to sustain a high localization accuracy. An experimental evaluation shows that SnapLoc exhibits a 90% error and median error of 33 cm and 18 cm, respectively, hence enabling decimeter-level accuracy at fast update rates for countless tags.
收起
摘要 :
A large body of work has shown that ultra-wideband (UWB) technology enables accurate indoor localization and tracking thanks to its high time-domain resolution. Existing systems, however, are typically designed to localize only a ...
展开
A large body of work has shown that ultra-wideband (UWB) technology enables accurate indoor localization and tracking thanks to its high time-domain resolution. Existing systems, however, are typically designed to localize only a limited number of tags, and involve the exchange of several messages following a given schedule. As a result, the scalability of current solutions in terms of tag density is limited, as well as their efficiency and responsiveness. In this paper, we present SnapLoc, a UWB-based indoor localization system that allows an unlimited number of tags to self-localize at a theoretical upper bound of 2.3 kHz. In SnapLoc, a tag obtains the responses from multiple anchors simultaneously. Based on these signals, the tag derives the time difference of arrival between anchors and estimates its position. Therefore, SnapLoc does not require tags to actively transmit packets, but to receive only a single message. This allows tags to passively localize themselves and ensures that the performance of SnapLoc does not degrade with high node densities. Moreover, due to the (quasi-)simultaneous responses, a tight clock synchronization between anchors is not needed. We have implemented SnapLoc on a low-cost platform based on the De-cawave DW1000 radio and solved limitations in the transceiver's timestamp resolution to sustain a high localization accuracy. An experimental evaluation shows that SnapLoc exhibits a 90% error and median error of 33 cm and 18 cm, respectively, hence enabling decimeter-level accuracy at fast update rates for countless tags.
收起
原文获取