Pub Date : 2018-09-01DOI: 10.1109/FAS-W.2018.00044
Henner Heck, B. Sick, Sven Tomforde
Self-improving system integration (SISSY) has been proposed as an approach to master Interwoven Systems and the resulting complexity issues. This results in a further transfer of traditional design-time decisions to runtime and from the engineer to the systems themselves. As a side-effect of the desired automatic reactions regarding the own integration status in an overall system structure, novel attack vectors and security threats appear. In this article, we classify the underlying attacker types, identify the SISSY-specific security threats, and propose research directions for counter measures to cope with these threats.
{"title":"Security Issues in Self-Improving System Integration – Challenges and Solution Strategies","authors":"Henner Heck, B. Sick, Sven Tomforde","doi":"10.1109/FAS-W.2018.00044","DOIUrl":"https://doi.org/10.1109/FAS-W.2018.00044","url":null,"abstract":"Self-improving system integration (SISSY) has been proposed as an approach to master Interwoven Systems and the resulting complexity issues. This results in a further transfer of traditional design-time decisions to runtime and from the engineer to the systems themselves. As a side-effect of the desired automatic reactions regarding the own integration status in an overall system structure, novel attack vectors and security threats appear. In this article, we classify the underlying attacker types, identify the SISSY-specific security threats, and propose research directions for counter measures to cope with these threats.","PeriodicalId":164903,"journal":{"name":"2018 IEEE 3rd International Workshops on Foundations and Applications of Self* Systems (FAS*W)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125279559","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-09-01DOI: 10.1109/FAS-W.2018.00053
Phyllis R. Nelson
Creation of large systems beyond systems of systems has begun. These interwoven systems (IwSs) create possibilities for extensive benefits, but also pose challenges to the safety of the critical infrastructure elements which are included. Encapsulation of the highest-value processes and functions of these critical infrastructures provides not only increased security, but also presents opportunities to develop more efficient means of discovery and description for use in the construction of IwSs. Encapsulation also presents opportunities to simplify the overall control challenges and the development of symbols and languages to support IwSs.
{"title":"Improving Security and Interoperability of Interwoven Systems through Rigorous Selective Encapsulation of Critical Physical Resources","authors":"Phyllis R. Nelson","doi":"10.1109/FAS-W.2018.00053","DOIUrl":"https://doi.org/10.1109/FAS-W.2018.00053","url":null,"abstract":"Creation of large systems beyond systems of systems has begun. These interwoven systems (IwSs) create possibilities for extensive benefits, but also pose challenges to the safety of the critical infrastructure elements which are included. Encapsulation of the highest-value processes and functions of these critical infrastructures provides not only increased security, but also presents opportunities to develop more efficient means of discovery and description for use in the construction of IwSs. Encapsulation also presents opportunities to simplify the overall control challenges and the development of symbols and languages to support IwSs.","PeriodicalId":164903,"journal":{"name":"2018 IEEE 3rd International Workshops on Foundations and Applications of Self* Systems (FAS*W)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117337692","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-09-01DOI: 10.1109/FAS-W.2018.00021
Jieun Choi, Geunchul Park, Dukyun Nam
Hardware performance counters in processors are mainly used for low level performance analysis and application tuning by monitoring performance-related hardware events. With the advent of processors with more cores than existing multicore processors and additional high-bandwidth memory, research on the performance analysis of new systems has received increasing attention from the high-performance computing community. Analyzing application characteristics and system features in a new system is essential for computational scientists and engineers who are eager to obtain the best performance of their scientific applications. However, these processors, increased core counts and high-performance resources, make it difficult to understand the correlation between performance-related hardware events. In this paper, we propose a method to simply and quickly classify application characteristics by using a data mining tool without understanding the correlation between hardware events. When we applied the proposed method to NAS Parallel Benchmarks (NPB), the application characteristics were the same as the authorized NPB categories. We show the effectiveness of the proposed scheme in a case study on analyzing the degree of interference between application characteristics.
{"title":"Efficient Classification of Application Characteristics by Using Hardware Performance Counters with Data Mining","authors":"Jieun Choi, Geunchul Park, Dukyun Nam","doi":"10.1109/FAS-W.2018.00021","DOIUrl":"https://doi.org/10.1109/FAS-W.2018.00021","url":null,"abstract":"Hardware performance counters in processors are mainly used for low level performance analysis and application tuning by monitoring performance-related hardware events. With the advent of processors with more cores than existing multicore processors and additional high-bandwidth memory, research on the performance analysis of new systems has received increasing attention from the high-performance computing community. Analyzing application characteristics and system features in a new system is essential for computational scientists and engineers who are eager to obtain the best performance of their scientific applications. However, these processors, increased core counts and high-performance resources, make it difficult to understand the correlation between performance-related hardware events. In this paper, we propose a method to simply and quickly classify application characteristics by using a data mining tool without understanding the correlation between hardware events. When we applied the proposed method to NAS Parallel Benchmarks (NPB), the application characteristics were the same as the authorized NPB categories. We show the effectiveness of the proposed scheme in a case study on analyzing the degree of interference between application characteristics.","PeriodicalId":164903,"journal":{"name":"2018 IEEE 3rd International Workshops on Foundations and Applications of Self* Systems (FAS*W)","volume":"285 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116232840","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-09-01DOI: 10.1109/FAS-W.2018.00025
Younghun Park, Minwoo Gu, Sun-Mi Yoo, Youngjae Kim, Sungyong Park
gVirt is a full GPU virtualization technique for Intel's integrated GPUs that alleviates the problems of other GPU virtualization techniques such as API remoting and direct pass-through. The original gVirt is known to have an inherent scalability limitation on the number of simultaneous virtual machines (VM). gScale solved this problem by allowing each VM to share a global graphics memory space and copy the entries in a private graphics translation table (GTT) to a physical GTT along with a GPU context switch. However, it still suffers from a large overhead of copying entries between private GTT and physical GTT, which becomes worse when the global graphics memory space allocated for each VM is overlapped. In this paper, we identify that the copy overhead caused by GPU context switch is the major bottleneck in performance improvement and propose a dynamic memory management scheme, called DymGPU, that provides two memory allocation algorithms such as size-based and utilization-based algorithms. While the size-based algorithm allocates memory space based on the memory size required by each VM, the utilization-based algorithm considers GPU utilization of each VM to allocate the memory space. DymGPU is also dynamic in the sense that the global graphics memory space used by each VM is rearranged at runtime by periodically checking idle VMs and GPU utilization of each runnable VM. We have implemented our proposed approach in gVirt and confirmed that the proposed scheme reduces GPU context switch time by up to 53% and improved the overall performance of various GPU applications by up to 39%.
{"title":"DymGPU: Dynamic Memory Management for Sharing GPUs in Virtualized Clouds","authors":"Younghun Park, Minwoo Gu, Sun-Mi Yoo, Youngjae Kim, Sungyong Park","doi":"10.1109/FAS-W.2018.00025","DOIUrl":"https://doi.org/10.1109/FAS-W.2018.00025","url":null,"abstract":"gVirt is a full GPU virtualization technique for Intel's integrated GPUs that alleviates the problems of other GPU virtualization techniques such as API remoting and direct pass-through. The original gVirt is known to have an inherent scalability limitation on the number of simultaneous virtual machines (VM). gScale solved this problem by allowing each VM to share a global graphics memory space and copy the entries in a private graphics translation table (GTT) to a physical GTT along with a GPU context switch. However, it still suffers from a large overhead of copying entries between private GTT and physical GTT, which becomes worse when the global graphics memory space allocated for each VM is overlapped. In this paper, we identify that the copy overhead caused by GPU context switch is the major bottleneck in performance improvement and propose a dynamic memory management scheme, called DymGPU, that provides two memory allocation algorithms such as size-based and utilization-based algorithms. While the size-based algorithm allocates memory space based on the memory size required by each VM, the utilization-based algorithm considers GPU utilization of each VM to allocate the memory space. DymGPU is also dynamic in the sense that the global graphics memory space used by each VM is rearranged at runtime by periodically checking idle VMs and GPU utilization of each runnable VM. We have implemented our proposed approach in gVirt and confirmed that the proposed scheme reduces GPU context switch time by up to 53% and improved the overall performance of various GPU applications by up to 39%.","PeriodicalId":164903,"journal":{"name":"2018 IEEE 3rd International Workshops on Foundations and Applications of Self* Systems (FAS*W)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114102638","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-09-01DOI: 10.1109/FAS-W.2018.00033
Roberto Casadei, Mirko Viroli
On the way to the materialisation of the pervasive computing vision, the technological progress swelling from mobile computing and the Internet of Things (IoT) domain is already rich of missed opportunities. Firstly, coordinating large numbers of heterogeneous situated entities to achieve system-level goals in a resilient and self-adaptive way is complex and requires novel approaches to be seamlessly injected into mainstream distributed computing models. Secondly, achieving effective exploitation of computer resources is difficult, due to operational constraints resulting from current paradigms and uncomprehensive software infrastructures which hinder flexibility, adaptation, and smooth coordination of computational tasks execution. Indeed, building dynamic, context-oriented applications in small-or large-scale IoT with traditional abstractions is hard: even harder is to achieve opportunistic, QoS-and QoE-driven application task management across available hardware and networking infrastructure. In this insight paper, we analyse by the collective adaptation perspective the key directions of the impelling paradigm shift urged by forthcoming large-scale IoT scenarios. Specifically, we consider how collective abstractions and platforms can synergistically assist in such a transformation, by better capturing and enacting a notion of "collective service" as well as the dynamic, opportunistic, and context-driven traits of space-time-situated computations.
{"title":"Collective Abstractions and Platforms for Large-Scale Self-Adaptive IoT","authors":"Roberto Casadei, Mirko Viroli","doi":"10.1109/FAS-W.2018.00033","DOIUrl":"https://doi.org/10.1109/FAS-W.2018.00033","url":null,"abstract":"On the way to the materialisation of the pervasive computing vision, the technological progress swelling from mobile computing and the Internet of Things (IoT) domain is already rich of missed opportunities. Firstly, coordinating large numbers of heterogeneous situated entities to achieve system-level goals in a resilient and self-adaptive way is complex and requires novel approaches to be seamlessly injected into mainstream distributed computing models. Secondly, achieving effective exploitation of computer resources is difficult, due to operational constraints resulting from current paradigms and uncomprehensive software infrastructures which hinder flexibility, adaptation, and smooth coordination of computational tasks execution. Indeed, building dynamic, context-oriented applications in small-or large-scale IoT with traditional abstractions is hard: even harder is to achieve opportunistic, QoS-and QoE-driven application task management across available hardware and networking infrastructure. In this insight paper, we analyse by the collective adaptation perspective the key directions of the impelling paradigm shift urged by forthcoming large-scale IoT scenarios. Specifically, we consider how collective abstractions and platforms can synergistically assist in such a transformation, by better capturing and enacting a notion of \"collective service\" as well as the dynamic, opportunistic, and context-driven traits of space-time-situated computations.","PeriodicalId":164903,"journal":{"name":"2018 IEEE 3rd International Workshops on Foundations and Applications of Self* Systems (FAS*W)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131140766","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-09-01DOI: 10.1109/FAS-W.2018.00052
K. Bellman
Before a self-integrating and self-improving (SISSY) system can become operational, we will have to demonstrate that some system properties and behaviors can be guaranteed to a reasonable degree. Reasonable means to the extent that, depending upon the application and functions of the system, we can use the system because we can guarantee some behaviors and properties to perform properly regardless of input, operational environment, or system goals. In this short paper, we discuss three classes of behaviors that can be guaranteed; these guarantees can be made through a combination of traditional pre-operational testing and confirmed with continual self-testing during operations. These can be briefly summarized as: (1) Fault detection, isolation, and recovery (FDIR) and top-down failure analysis; (2) Minimal self-protection; (3) Minimal acceptable performance.
{"title":"What Reasonable Guarantees Can We Make for a SISSY System","authors":"K. Bellman","doi":"10.1109/FAS-W.2018.00052","DOIUrl":"https://doi.org/10.1109/FAS-W.2018.00052","url":null,"abstract":"Before a self-integrating and self-improving (SISSY) system can become operational, we will have to demonstrate that some system properties and behaviors can be guaranteed to a reasonable degree. Reasonable means to the extent that, depending upon the application and functions of the system, we can use the system because we can guarantee some behaviors and properties to perform properly regardless of input, operational environment, or system goals. In this short paper, we discuss three classes of behaviors that can be guaranteed; these guarantees can be made through a combination of traditional pre-operational testing and confirmed with continual self-testing during operations. These can be briefly summarized as: (1) Fault detection, isolation, and recovery (FDIR) and top-down failure analysis; (2) Minimal self-protection; (3) Minimal acceptable performance.","PeriodicalId":164903,"journal":{"name":"2018 IEEE 3rd International Workshops on Foundations and Applications of Self* Systems (FAS*W)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133127753","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-09-01DOI: 10.1109/FAS-W.2018.00020
C. Raibulet
Self-* properties characterize dynamic software able to perform changes on itself by itself during its execution. The objectives of these changes are to maintain the functionality for which the software has been implemented and its related quality, and/or to improve the performances of the software whenever this is possible. Changes aim to address a wide range of issues, e.g., from resource variability (e.g., due to mobility) and changing users' needs to security threats and faults. One of the main advantages of self-* software is that the complexity of changes is managed dynamically by the software and hidden from the users. Today, more and more software solutions are characterized by self-* properties. The objective of this paper is to investigate which are the evaluation approaches of self-* software and how self-* properties may be evaluated. Further, a taxonomy for the evaluation of self-* software is proposed.
{"title":"Towards a Taxonomy for the Evaluation of Self-* Software","authors":"C. Raibulet","doi":"10.1109/FAS-W.2018.00020","DOIUrl":"https://doi.org/10.1109/FAS-W.2018.00020","url":null,"abstract":"Self-* properties characterize dynamic software able to perform changes on itself by itself during its execution. The objectives of these changes are to maintain the functionality for which the software has been implemented and its related quality, and/or to improve the performances of the software whenever this is possible. Changes aim to address a wide range of issues, e.g., from resource variability (e.g., due to mobility) and changing users' needs to security threats and faults. One of the main advantages of self-* software is that the complexity of changes is managed dynamically by the software and hidden from the users. Today, more and more software solutions are characterized by self-* properties. The objective of this paper is to investigate which are the evaluation approaches of self-* software and how self-* properties may be evaluated. Further, a taxonomy for the evaluation of self-* software is proposed.","PeriodicalId":164903,"journal":{"name":"2018 IEEE 3rd International Workshops on Foundations and Applications of Self* Systems (FAS*W)","volume":"90 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131028893","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-09-01DOI: 10.1109/FAS-W.2018.00038
J. Pitt, Rui P. Cardoso, E. Hart, Josiah Ober
Many applications of collective adaptive systems for the digital transformation or digital society will necessarily be multi-functional; that is, the collective, as it adapts over time, will be required to resolve many and different types of problem. However, a long-standing issue for political theorists has been whether a decentralised problem-solving regime can be both 'democratic' and 'epistemic', i.e. is it possible to devise decision-making and action-determination processes that take into account both majority preference and expert judgement. In this paper, we address this issue in the context of engineering long-lived and sustainable collective adaptive systems, in which autonomous agents adapt conventional rules in order to be congruent with changes in their operating environment. Based on a preliminary proof of concept and inspiration from political science, we propose a reference architecture for relevant expertise aggregation. We conclude that this is one possible design solution to the problem of enabling an collective to assume direct responsibility for adaptation or adoption of problem-solving policies at a large scale, over long periods of time, and addressing diverse problem types.
{"title":"Relevant Expertise Aggregation for Policy Selection in Collective Adaptive Systems","authors":"J. Pitt, Rui P. Cardoso, E. Hart, Josiah Ober","doi":"10.1109/FAS-W.2018.00038","DOIUrl":"https://doi.org/10.1109/FAS-W.2018.00038","url":null,"abstract":"Many applications of collective adaptive systems for the digital transformation or digital society will necessarily be multi-functional; that is, the collective, as it adapts over time, will be required to resolve many and different types of problem. However, a long-standing issue for political theorists has been whether a decentralised problem-solving regime can be both 'democratic' and 'epistemic', i.e. is it possible to devise decision-making and action-determination processes that take into account both majority preference and expert judgement. In this paper, we address this issue in the context of engineering long-lived and sustainable collective adaptive systems, in which autonomous agents adapt conventional rules in order to be congruent with changes in their operating environment. Based on a preliminary proof of concept and inspiration from political science, we propose a reference architecture for relevant expertise aggregation. We conclude that this is one possible design solution to the problem of enabling an collective to assume direct responsibility for adaptation or adoption of problem-solving policies at a large scale, over long periods of time, and addressing diverse problem types.","PeriodicalId":164903,"journal":{"name":"2018 IEEE 3rd International Workshops on Foundations and Applications of Self* Systems (FAS*W)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134120699","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-07-12DOI: 10.1109/FAS-W.2018.00043
A. Diaconescu, Barry Porter, Roberto Rodrigues Filho, Evangelos Pournaras
System self-integration from open sets of components provides the basis for open adaptability to unpredictable environments. Hierarchical architectures are essential for enabling such systems to scale, as they allow to compromise between processing detailed knowledge in parallel and coordinating parallel processes from a more abstract viewpoint; recursively. This position paper aims to bring to the fore the following key design aspect of such hierarchical systems: how should the authority of decision and action be assigned across hierarchical levels, with respect to the self-awareness capabilities of these levels, The difficulty lays in that all levels lack knowledge, which may be key to certain decisions, because lower levels have detailed knowledge but within a narrow scope (good for local customisation), and higher levels have a broader scope but no details (good for global coordination). We highlight the most obvious authority schemes available and discuss their advantages and shortcomings: top-down, bottom-up, and iterative (yoyo). We discuss three detailed application examples from our previous work on hierarchical systems, pointing-out the knowledge and authority schemes employed and the possible alternatives. This provides a basis for offering system designers the necessary understanding and tools for taking the appropriate decisions with respect to the distribution of self-awareness capabilities and authority of decision and action across hierarchical system levels.
{"title":"Hierarchical Self-Awareness and Authority for Scalable Self-Integrating Systems","authors":"A. Diaconescu, Barry Porter, Roberto Rodrigues Filho, Evangelos Pournaras","doi":"10.1109/FAS-W.2018.00043","DOIUrl":"https://doi.org/10.1109/FAS-W.2018.00043","url":null,"abstract":"System self-integration from open sets of components provides the basis for open adaptability to unpredictable environments. Hierarchical architectures are essential for enabling such systems to scale, as they allow to compromise between processing detailed knowledge in parallel and coordinating parallel processes from a more abstract viewpoint; recursively. This position paper aims to bring to the fore the following key design aspect of such hierarchical systems: how should the authority of decision and action be assigned across hierarchical levels, with respect to the self-awareness capabilities of these levels, The difficulty lays in that all levels lack knowledge, which may be key to certain decisions, because lower levels have detailed knowledge but within a narrow scope (good for local customisation), and higher levels have a broader scope but no details (good for global coordination). We highlight the most obvious authority schemes available and discuss their advantages and shortcomings: top-down, bottom-up, and iterative (yoyo). We discuss three detailed application examples from our previous work on hierarchical systems, pointing-out the knowledge and authority schemes employed and the possible alternatives. This provides a basis for offering system designers the necessary understanding and tools for taking the appropriate decisions with respect to the distribution of self-awareness capabilities and authority of decision and action across hierarchical system levels.","PeriodicalId":164903,"journal":{"name":"2018 IEEE 3rd International Workshops on Foundations and Applications of Self* Systems (FAS*W)","volume":"330 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-07-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134332615","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2018-07-03DOI: 10.1109/FAS-W.2018.00046
Stephan Deist, Maarten Bieshaar, Jens Schreiber, André Gensler, B. Sick
In this article, we propose the Coopetititve Soft Gating Ensemble or CSGE for general machine learning tasks and interwoven systems.The goal of machine learning is to create models that generalize well for unknown datasets. Often, however, the problems are too complex to be solved with a single model, so several models are combined. Similar, Autonomic Computing requires the integration of different systems. Here, especially, the local, temporal online evaluation and the resulting (re-)weighting scheme of the CSGE makes the approach highly applicable for self-improving system integrations. To achieve the best potential performance the CSGE can be optimized according to arbitrary loss functions making it accessible for a broader range of problems. We introduce a novel training procedure including a hyper-parameter initialisation at its heart. We show that the CSGE approach reaches state-of-the-art performance for both classification and regression tasks. Further on, the CSGE provides a human-readable quantification on the influence of all base estimators employing the three weighting aspects. Moreover, we provide a scikit-learn compatible implementation.
{"title":"Coopetitive Soft Gating Ensemble","authors":"Stephan Deist, Maarten Bieshaar, Jens Schreiber, André Gensler, B. Sick","doi":"10.1109/FAS-W.2018.00046","DOIUrl":"https://doi.org/10.1109/FAS-W.2018.00046","url":null,"abstract":"In this article, we propose the Coopetititve Soft Gating Ensemble or CSGE for general machine learning tasks and interwoven systems.The goal of machine learning is to create models that generalize well for unknown datasets. Often, however, the problems are too complex to be solved with a single model, so several models are combined. Similar, Autonomic Computing requires the integration of different systems. Here, especially, the local, temporal online evaluation and the resulting (re-)weighting scheme of the CSGE makes the approach highly applicable for self-improving system integrations. To achieve the best potential performance the CSGE can be optimized according to arbitrary loss functions making it accessible for a broader range of problems. We introduce a novel training procedure including a hyper-parameter initialisation at its heart. We show that the CSGE approach reaches state-of-the-art performance for both classification and regression tasks. Further on, the CSGE provides a human-readable quantification on the influence of all base estimators employing the three weighting aspects. Moreover, we provide a scikit-learn compatible implementation.","PeriodicalId":164903,"journal":{"name":"2018 IEEE 3rd International Workshops on Foundations and Applications of Self* Systems (FAS*W)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-07-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124871542","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：