Pub Date : 2016-12-01DOI: 10.1109/SSCI.2016.7850094
Yue Dai, Yuan Ma, Qianyi Wang, Y. Murphey, Shiqi Qiu, Johannes Kristinsson, Jason Meyer, F. Tseng, T. Feldkamp
Personal route prediction (PRP) has attracted much research interest recently because of its technical challenges and broad applications in intelligent vehicle and transportation systems. Traditional navigation systems generate a route for a given origin and destination based on either shortest or fastest route schemes. In practice, different people may very likely take different routes from the same origin to the same destination. Personal route prediction attempts to predict a driver's route based on the knowledge of driver's preferences. In this paper we present an intelligent personal route prediction system, I_PRP, which is built based upon a knowledge base of personal route preference learned from driver's historical trips. The I_PRP contains an intelligent route prediction algorithm based on the first order Markov chain model to predict a driver's intended route for a given pair of origin and destination, and a dynamic route prediction algorithm that has the capability of predicting driver's new route after the driver departs from the predicted route.
{"title":"Dynamic prediction of drivers' personal routes through machine learning","authors":"Yue Dai, Yuan Ma, Qianyi Wang, Y. Murphey, Shiqi Qiu, Johannes Kristinsson, Jason Meyer, F. Tseng, T. Feldkamp","doi":"10.1109/SSCI.2016.7850094","DOIUrl":"https://doi.org/10.1109/SSCI.2016.7850094","url":null,"abstract":"Personal route prediction (PRP) has attracted much research interest recently because of its technical challenges and broad applications in intelligent vehicle and transportation systems. Traditional navigation systems generate a route for a given origin and destination based on either shortest or fastest route schemes. In practice, different people may very likely take different routes from the same origin to the same destination. Personal route prediction attempts to predict a driver's route based on the knowledge of driver's preferences. In this paper we present an intelligent personal route prediction system, I_PRP, which is built based upon a knowledge base of personal route preference learned from driver's historical trips. The I_PRP contains an intelligent route prediction algorithm based on the first order Markov chain model to predict a driver's intended route for a given pair of origin and destination, and a dynamic route prediction algorithm that has the capability of predicting driver's new route after the driver departs from the predicted route.","PeriodicalId":120288,"journal":{"name":"2016 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"9 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133724675","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 : 2016-12-01DOI: 10.1109/SSCI.2016.7850060
A. Serbencu, V. Mînzu
Ant Colony System is a well-known metaheuristic used to solve combinatorial optimization problems that is not intrinsically prepared to deal with precedence constraints. The work reported here is the continuation of the results presented in a previous paper that proposed an Ant System algorithm devoted to Tasks to Workstations Assignment problem. A special technique was developed in order to increase the effectiveness of precedence constraints treatment. On the one hand the contribution of this paper consists in the amelioration of this technique. On the other hand, the Ant System algorithm is hybridized with a local descent deterministic algorithm that contributes greatly to the avoiding of solutions bias. The results of the hybridized Ant System algorithm have proved the effectiveness of the proposed way to treat the precedence constraints
{"title":"Hybridized Ant Colony System for Tasks to Workstations Assignment","authors":"A. Serbencu, V. Mînzu","doi":"10.1109/SSCI.2016.7850060","DOIUrl":"https://doi.org/10.1109/SSCI.2016.7850060","url":null,"abstract":"Ant Colony System is a well-known metaheuristic used to solve combinatorial optimization problems that is not intrinsically prepared to deal with precedence constraints. The work reported here is the continuation of the results presented in a previous paper that proposed an Ant System algorithm devoted to Tasks to Workstations Assignment problem. A special technique was developed in order to increase the effectiveness of precedence constraints treatment. On the one hand the contribution of this paper consists in the amelioration of this technique. On the other hand, the Ant System algorithm is hybridized with a local descent deterministic algorithm that contributes greatly to the avoiding of solutions bias. The results of the hybridized Ant System algorithm have proved the effectiveness of the proposed way to treat the precedence constraints","PeriodicalId":120288,"journal":{"name":"2016 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115451703","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 : 2016-12-01DOI: 10.1109/SSCI.2016.7849951
K. D. Gupta, Renuka Pampana, R. Vilalta, E. Ishida, R. S. Souza
While the current supernova (SN) photometric classification system is based on high resolution spectroscopic observations, the next generation of large scale surveys will be based on photometric light curves of supernovae gathered at an unprecedented rate. Developing an efficient method for SN photometric classification is critical to cope with the rapid growth of data volumes in current astronomical surveys. In this work, we present an adaptive mechanism that generates a predictive model to identify a particular class of SN known as Type Ia, when the source set is made of spectroscopic data, while the target set is made of photometric data. The method is applied to simulated data sets derived from the Supernova Photometric Classification Challenge, and preprocessed using Gaussian Process Regression for all objects with at least 1 observational epoch before -3 and after +24 days since the SN maximum brightness. The main difficulty lies in the compatibility of models between spectroscopic (source) data and photometric (target) data, since the underlying distributions on both, source and target domains, are expected to be significantly different. A solution is to adapt predictive models across domains. Our methodology exploits machine learning techniques by combining two concepts: 1) domain adaptation is used to transfer properties from the source domain to the target domain; and 2) active learning is used as a means to rely on a set of confident labels on the target domain. We show how a combination of both concepts leads to high generalization (i.e., predictive) performance.
{"title":"Automated supernova Ia classification using adaptive learning techniques","authors":"K. D. Gupta, Renuka Pampana, R. Vilalta, E. Ishida, R. S. Souza","doi":"10.1109/SSCI.2016.7849951","DOIUrl":"https://doi.org/10.1109/SSCI.2016.7849951","url":null,"abstract":"While the current supernova (SN) photometric classification system is based on high resolution spectroscopic observations, the next generation of large scale surveys will be based on photometric light curves of supernovae gathered at an unprecedented rate. Developing an efficient method for SN photometric classification is critical to cope with the rapid growth of data volumes in current astronomical surveys. In this work, we present an adaptive mechanism that generates a predictive model to identify a particular class of SN known as Type Ia, when the source set is made of spectroscopic data, while the target set is made of photometric data. The method is applied to simulated data sets derived from the Supernova Photometric Classification Challenge, and preprocessed using Gaussian Process Regression for all objects with at least 1 observational epoch before -3 and after +24 days since the SN maximum brightness. The main difficulty lies in the compatibility of models between spectroscopic (source) data and photometric (target) data, since the underlying distributions on both, source and target domains, are expected to be significantly different. A solution is to adapt predictive models across domains. Our methodology exploits machine learning techniques by combining two concepts: 1) domain adaptation is used to transfer properties from the source domain to the target domain; and 2) active learning is used as a means to rely on a set of confident labels on the target domain. We show how a combination of both concepts leads to high generalization (i.e., predictive) performance.","PeriodicalId":120288,"journal":{"name":"2016 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"90 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115538290","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 : 2016-12-01DOI: 10.1109/SSCI.2016.7849850
Denise M. Case, C. Stylios
The power market is becoming more complex as independent small producers are entering it but their energy offerings are often based on alternative sources which may be dependent on transient weather conditions. Power market auction behavior is a typical large-scale system characterized by huge amounts of data and information that have to be taken into consideration to make decisions. Fuzzy Cognitive Maps (FCM) offer a method for using the knowledge and experience of domain experts to describe the behavior of a complex system. This paper discusses FCM representation and development, and describes the use of FCM to develop a behavioral model of the system. This paper then presents the soft computing approach of FCM for modeling complex power market behavior. The resulting FCM models a variety of factors that affect individual participant behaviors during power auctions and provides an abstract conceptual model of the interacting entities for a specific case problem.
{"title":"Introducing a Fuzzy Cognitive Map for modeling power market auction behavior","authors":"Denise M. Case, C. Stylios","doi":"10.1109/SSCI.2016.7849850","DOIUrl":"https://doi.org/10.1109/SSCI.2016.7849850","url":null,"abstract":"The power market is becoming more complex as independent small producers are entering it but their energy offerings are often based on alternative sources which may be dependent on transient weather conditions. Power market auction behavior is a typical large-scale system characterized by huge amounts of data and information that have to be taken into consideration to make decisions. Fuzzy Cognitive Maps (FCM) offer a method for using the knowledge and experience of domain experts to describe the behavior of a complex system. This paper discusses FCM representation and development, and describes the use of FCM to develop a behavioral model of the system. This paper then presents the soft computing approach of FCM for modeling complex power market behavior. The resulting FCM models a variety of factors that affect individual participant behaviors during power auctions and provides an abstract conceptual model of the interacting entities for a specific case problem.","PeriodicalId":120288,"journal":{"name":"2016 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115901561","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 : 2016-12-01DOI: 10.1109/SSCI.2016.7850142
Xiao-Bing Hu, Ming-Kong Zhang, Jian-Qin Liao
To assess the performance of a network system against disturbances, existing methods are usually concerned with two different extreme situations: (i) how likely the system will degrade into some separated sub-graphs because of disturbances; (ii) how likely the 1st best paths will be cut off by disturbances. However, a more general situation, i.e., how likely those paths whose lengths are within a given range will be affected by disturbances, is barely discussed. Basically, to address this general situation, all (not just a proportion) of those paths whose lengths are within the given range must be found out between all pairs of origin and destination (OD) of interest. Unfortunately, no effective method has ever been reported to accomplish this task, although there are many methods capable of calculating all the 1st best paths between all OD pair of interest. This paper, for the first time, attempts to address the above general situation of network performance assessment. To this end, a novel ripple-spreading algorithm (RSA) is proposed to carry out a ripple relay race on the network, in order to identify all of those paths whose lengths are within the given range. Surprisingly, the proposed RSA can find all such paths between all OD pairs of interest by just a single run of ripple relay race. This work makes progress towards the general performance assessment of a network system against disturbances.
{"title":"A ripple-spreading algorithm for network performance assessment","authors":"Xiao-Bing Hu, Ming-Kong Zhang, Jian-Qin Liao","doi":"10.1109/SSCI.2016.7850142","DOIUrl":"https://doi.org/10.1109/SSCI.2016.7850142","url":null,"abstract":"To assess the performance of a network system against disturbances, existing methods are usually concerned with two different extreme situations: (i) how likely the system will degrade into some separated sub-graphs because of disturbances; (ii) how likely the 1st best paths will be cut off by disturbances. However, a more general situation, i.e., how likely those paths whose lengths are within a given range will be affected by disturbances, is barely discussed. Basically, to address this general situation, all (not just a proportion) of those paths whose lengths are within the given range must be found out between all pairs of origin and destination (OD) of interest. Unfortunately, no effective method has ever been reported to accomplish this task, although there are many methods capable of calculating all the 1st best paths between all OD pair of interest. This paper, for the first time, attempts to address the above general situation of network performance assessment. To this end, a novel ripple-spreading algorithm (RSA) is proposed to carry out a ripple relay race on the network, in order to identify all of those paths whose lengths are within the given range. Surprisingly, the proposed RSA can find all such paths between all OD pairs of interest by just a single run of ripple relay race. This work makes progress towards the general performance assessment of a network system against disturbances.","PeriodicalId":120288,"journal":{"name":"2016 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"122 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124382830","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 : 2016-12-01DOI: 10.1109/SSCI.2016.7849906
Hanan Hibshi, T. Breaux, Christian Wagner
Organizations rely on security experts to improve the security of their systems. These professionals use background knowledge and experience to align known threats and vulnerabilities before selecting mitigation options. The substantial depth of expertise in any one area (e.g., databases, networks, operating systems) precludes the possibility that an expert would have complete knowledge about all threats and vulnerabilities. To begin addressing this problem of fragmented knowledge, we investigate the challenge of developing a security requirements rule base that mimics multi-human expert reasoning to enable new decision-support systems. In this paper, we show how to collect relevant information from cyber security experts to enable the generation of: (1) interval type-2 fuzzy sets that capture intra- and inter-expert uncertainty around vulnerability levels; and (2) fuzzy logic rules driving the decision-making process within the requirements analysis. The proposed method relies on comparative ratings of security requirements in the context of concrete vignettes, providing a novel, interdisciplinary approach to knowledge generation for fuzzy logic systems. The paper presents an initial evaluation of the proposed approach through 52 scenarios with 13 experts to compare their assessments to those of the fuzzy logic decision support system. The results show that the system provides reliable assessments to the security analysts, in particular, generating more conservative assessments in 19% of the test scenarios compared to the experts' ratings.
{"title":"Improving security requirements adequacy","authors":"Hanan Hibshi, T. Breaux, Christian Wagner","doi":"10.1109/SSCI.2016.7849906","DOIUrl":"https://doi.org/10.1109/SSCI.2016.7849906","url":null,"abstract":"Organizations rely on security experts to improve the security of their systems. These professionals use background knowledge and experience to align known threats and vulnerabilities before selecting mitigation options. The substantial depth of expertise in any one area (e.g., databases, networks, operating systems) precludes the possibility that an expert would have complete knowledge about all threats and vulnerabilities. To begin addressing this problem of fragmented knowledge, we investigate the challenge of developing a security requirements rule base that mimics multi-human expert reasoning to enable new decision-support systems. In this paper, we show how to collect relevant information from cyber security experts to enable the generation of: (1) interval type-2 fuzzy sets that capture intra- and inter-expert uncertainty around vulnerability levels; and (2) fuzzy logic rules driving the decision-making process within the requirements analysis. The proposed method relies on comparative ratings of security requirements in the context of concrete vignettes, providing a novel, interdisciplinary approach to knowledge generation for fuzzy logic systems. The paper presents an initial evaluation of the proposed approach through 52 scenarios with 13 experts to compare their assessments to those of the fuzzy logic decision support system. The results show that the system provides reliable assessments to the security analysts, in particular, generating more conservative assessments in 19% of the test scenarios compared to the experts' ratings.","PeriodicalId":120288,"journal":{"name":"2016 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114438355","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 : 2016-12-01DOI: 10.1109/SSCI.2016.7850256
U. Mlakar, J. Brest, Iztok Fister, Iztok Fister
Image segmentation is an important preprocessing step in many computer vision applications, using the image thresholding as one of the simplest and the most applied methods. Since the optimal thresholds' selection can be regarded as an optimization problem, it can be found easily by applying any meta-heuristic with an appropriate objective function. This paper investigates the impact of different chaotic maps, embedded into a self-adaptive differential evolution for the purpose of image thresholding. The Kapur entropy is used as an objective function that maximizes the entropy of different regions in the image. Three chaotic maps, namely the Kent, Logistic and Tent, found commonly in literature, are studied in this paper. The applied chaotic maps are compared to the original differential evolution, self-adaptive differential evolution, and the state-of-the-art L-Shade tested on four images. The results show that the applied chaotic maps improve the results obtained using the traditional randomized method.
{"title":"A study of chaotic maps in differential evolution applied to gray-level image thresholding","authors":"U. Mlakar, J. Brest, Iztok Fister, Iztok Fister","doi":"10.1109/SSCI.2016.7850256","DOIUrl":"https://doi.org/10.1109/SSCI.2016.7850256","url":null,"abstract":"Image segmentation is an important preprocessing step in many computer vision applications, using the image thresholding as one of the simplest and the most applied methods. Since the optimal thresholds' selection can be regarded as an optimization problem, it can be found easily by applying any meta-heuristic with an appropriate objective function. This paper investigates the impact of different chaotic maps, embedded into a self-adaptive differential evolution for the purpose of image thresholding. The Kapur entropy is used as an objective function that maximizes the entropy of different regions in the image. Three chaotic maps, namely the Kent, Logistic and Tent, found commonly in literature, are studied in this paper. The applied chaotic maps are compared to the original differential evolution, self-adaptive differential evolution, and the state-of-the-art L-Shade tested on four images. The results show that the applied chaotic maps improve the results obtained using the traditional randomized method.","PeriodicalId":120288,"journal":{"name":"2016 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115089404","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 : 2016-12-01DOI: 10.1109/SSCI.2016.7849964
J. A. Brown, D. Ashlock
This study presents an application of the Multiple Worlds Model of Evolution. The goal is to model radio stations in a given market. The model captures listener demographics and maximizes listeners, while securing advertising revenue. Listener preferences for different types of content are set as positive (like) and negative (dislike) integers, allowing surveys of the demographic to act as the model parameters directly. Fitness evaluation is performed with a modeled hour of radio playtime where stations can select between a set of content types and advertisements. Advertisements provide fitness in the form of advertising revenues; however, listeners will only stay on a station which provides content they enjoy. The Multiple Worlds Model is a form of multiple population evolutionary algorithm. It evaluates fitness based on the actions of one member from each population, and has no genetic transfer of information between populations. Each population can thus specialize. In the current study, such specialization is a self-organization of focused (e.g. rock or country) stations via adaption to listener preferences. The model is examined using different numbers of independent populations with even splits among demographic types. The evolved stations show differences in playlists where the profiles differ in their enjoyments and convergence between stations where the listener profiles are similar.
{"title":"Multiple Worlds Model of Evolution for demographic appropriate radio playlists","authors":"J. A. Brown, D. Ashlock","doi":"10.1109/SSCI.2016.7849964","DOIUrl":"https://doi.org/10.1109/SSCI.2016.7849964","url":null,"abstract":"This study presents an application of the Multiple Worlds Model of Evolution. The goal is to model radio stations in a given market. The model captures listener demographics and maximizes listeners, while securing advertising revenue. Listener preferences for different types of content are set as positive (like) and negative (dislike) integers, allowing surveys of the demographic to act as the model parameters directly. Fitness evaluation is performed with a modeled hour of radio playtime where stations can select between a set of content types and advertisements. Advertisements provide fitness in the form of advertising revenues; however, listeners will only stay on a station which provides content they enjoy. The Multiple Worlds Model is a form of multiple population evolutionary algorithm. It evaluates fitness based on the actions of one member from each population, and has no genetic transfer of information between populations. Each population can thus specialize. In the current study, such specialization is a self-organization of focused (e.g. rock or country) stations via adaption to listener preferences. The model is examined using different numbers of independent populations with even splits among demographic types. The evolved stations show differences in playlists where the profiles differ in their enjoyments and convergence between stations where the listener profiles are similar.","PeriodicalId":120288,"journal":{"name":"2016 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"72 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123215252","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 : 2016-12-01DOI: 10.1109/SSCI.2016.7850152
Anna Sergeevna Bosman, A. Engelbrecht, Mardé Helbig
Fitness landscape analysis encompasses a selection of techniques designed to estimate the properties of a search landscape associated with an optimisation problem. Applied to neural network training, fitness landscape analysis can be used to establish the link between the shape of the objective function and various neural network design and architecture properties. However, most fitness landscape analysis metrics rely on search space sampling. Since neural network search space is unbounded, it is unclear what subset of the search space should be sampled to obtain representative measurements. This study analyses fitness landscape properties of neural networks under various search space boundaries, and proposes meaningful search space bounds for neural network fitness landscape analysis.
{"title":"Search space boundaries in neural network error landscape analysis","authors":"Anna Sergeevna Bosman, A. Engelbrecht, Mardé Helbig","doi":"10.1109/SSCI.2016.7850152","DOIUrl":"https://doi.org/10.1109/SSCI.2016.7850152","url":null,"abstract":"Fitness landscape analysis encompasses a selection of techniques designed to estimate the properties of a search landscape associated with an optimisation problem. Applied to neural network training, fitness landscape analysis can be used to establish the link between the shape of the objective function and various neural network design and architecture properties. However, most fitness landscape analysis metrics rely on search space sampling. Since neural network search space is unbounded, it is unclear what subset of the search space should be sampled to obtain representative measurements. This study analyses fitness landscape properties of neural networks under various search space boundaries, and proposes meaningful search space bounds for neural network fitness landscape analysis.","PeriodicalId":120288,"journal":{"name":"2016 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"191 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123483377","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 : 2016-12-01DOI: 10.1109/SSCI.2016.7849974
M. Sommer, Anthony Stein, J. Hähner
Time series forecasting constitutes an important aspect of any kind of technical system, since the underlying stochastic processes vary over time. Extensive efforts for designing self-adaptive learning systems have been made, to take system designers out of the loop. One goal of such systems is to transfer design-time decisions, e.g. parametrisation, to the run-time. By means of forecasting the succeeding system state, the system itself is enabled to anticipate, how to reconfigure to handle upcoming conditions. Ensemble forecasting is a specific means of combining and weighting the forecasts of multiple independent forecast methods. This concept has proven successful in various domains today. In this work, we present our self-adaptive forecast module for ensemble forecasting of univariate time series and draw a picture of how the eXtended Classifier System for Function approximation (XCSF) can be utilised as a novel weighting approach in this context. We elaborate on the fundamental ideas and evaluate our proposed technique on the basis of several time series with different characteristics.
{"title":"Local ensemble weighting in the context of time series forecasting using XCSF","authors":"M. Sommer, Anthony Stein, J. Hähner","doi":"10.1109/SSCI.2016.7849974","DOIUrl":"https://doi.org/10.1109/SSCI.2016.7849974","url":null,"abstract":"Time series forecasting constitutes an important aspect of any kind of technical system, since the underlying stochastic processes vary over time. Extensive efforts for designing self-adaptive learning systems have been made, to take system designers out of the loop. One goal of such systems is to transfer design-time decisions, e.g. parametrisation, to the run-time. By means of forecasting the succeeding system state, the system itself is enabled to anticipate, how to reconfigure to handle upcoming conditions. Ensemble forecasting is a specific means of combining and weighting the forecasts of multiple independent forecast methods. This concept has proven successful in various domains today. In this work, we present our self-adaptive forecast module for ensemble forecasting of univariate time series and draw a picture of how the eXtended Classifier System for Function approximation (XCSF) can be utilised as a novel weighting approach in this context. We elaborate on the fundamental ideas and evaluate our proposed technique on the basis of several time series with different characteristics.","PeriodicalId":120288,"journal":{"name":"2016 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"64 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122007971","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：