Pub Date : 2019-12-01DOI: 10.1109/SSCI44817.2019.9002943
Xiangping Chen, W. Cao, Lei Xing
Utilization of renewable energy (e.g. wind, solar, bio-energy) is high on the governmental agenda globally. In order to tackle energy poverty and increase energy efficiency in energy systems, a hybrid energy system including wind, hydrogen and fuel cells is proposed to supplement to the main power grid. Wind energy is firstly converted into electrical energy while part of the generated electricity is used for water electrolysis to generate hydrogen for energy storage. Hydrogen is used by fuel cells to convert to electricity when electrical energy demand peaks. An analytical model is developed to coordinate the operation of the system involving energy conversion between hydrogen, electrical and mechanical forms. The proposed system is primarily designed to meet the electrical demand of a rural village while the energy storage system can meet the discrepancy between intermittent renewable energy supplies and fluctuated energy demands so as to improve the system efficiency. Genetic Algorithm (GA) is used as an optimization strategy to determine the operational scheme for the multi-vector energy system. In this work, case studies are carried out based on actual measurement data. The test results have confirmed the effectiveness of the proposed methodology and maximizing the wind energy consumption locally. This is an alternative to battery energy storage and can be widely used in wind-rich rural areas.
{"title":"GA-Aided Power Flow Management in a Multi-Vector Energy System","authors":"Xiangping Chen, W. Cao, Lei Xing","doi":"10.1109/SSCI44817.2019.9002943","DOIUrl":"https://doi.org/10.1109/SSCI44817.2019.9002943","url":null,"abstract":"Utilization of renewable energy (e.g. wind, solar, bio-energy) is high on the governmental agenda globally. In order to tackle energy poverty and increase energy efficiency in energy systems, a hybrid energy system including wind, hydrogen and fuel cells is proposed to supplement to the main power grid. Wind energy is firstly converted into electrical energy while part of the generated electricity is used for water electrolysis to generate hydrogen for energy storage. Hydrogen is used by fuel cells to convert to electricity when electrical energy demand peaks. An analytical model is developed to coordinate the operation of the system involving energy conversion between hydrogen, electrical and mechanical forms. The proposed system is primarily designed to meet the electrical demand of a rural village while the energy storage system can meet the discrepancy between intermittent renewable energy supplies and fluctuated energy demands so as to improve the system efficiency. Genetic Algorithm (GA) is used as an optimization strategy to determine the operational scheme for the multi-vector energy system. In this work, case studies are carried out based on actual measurement data. The test results have confirmed the effectiveness of the proposed methodology and maximizing the wind energy consumption locally. This is an alternative to battery energy storage and can be widely used in wind-rich rural areas.","PeriodicalId":6729,"journal":{"name":"2019 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"7 5 1","pages":"3172-3176"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77407774","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 : 2019-12-01DOI: 10.1109/SSCI44817.2019.9002750
Tushar Garg, S. Roy
This work proposes a consensus-based novel PI-like parameter estimator for collaborative system identification. Conventional online parameter estimation algorithms, which are used for system identification, require a restrictive condition of persistence of excitation (PE) for the estimates to converge to the true parameters. Some recent works have shown that collaborative system identification using multiple agents can relax the PE condition to a milder condition of collective persistence of excitation (C-PE) for parameter convergence. The C-PE condition implies that the PE condition is cooperatively satisfied by all the agents through sharing information between neighbors using a connected graph architecture, where each individual agent does not require to satisfy the PE condition separately. The proposed work designs a novel collaborative parameter estimator dynamics, which with the help of integral-like component ensures parameter convergence under a further slackened condition; coined as collective Initial Excitation (C-IE). The C-IE condition is an extension of the concept of initial excitation (IE), which is recently proposed in the context of parameter estimation in adaptive control. It has been already established that IE condition is significantly less restrictive than PE. The current work generalizes the concept of IE in a multi-agent settings, where information sharing through connected graph guarantees consensus parameter convergence under the C-IE condition. It can be argued that C-IE condition is milder than all of the other above mentioned conditions of PE, C-PE and IE. Simulation results further validate the efficacy of the proposed estimation algorithm.
{"title":"Collaborative System Identification via Consensus-Based novel PI-like Parameter Estimator","authors":"Tushar Garg, S. Roy","doi":"10.1109/SSCI44817.2019.9002750","DOIUrl":"https://doi.org/10.1109/SSCI44817.2019.9002750","url":null,"abstract":"This work proposes a consensus-based novel PI-like parameter estimator for collaborative system identification. Conventional online parameter estimation algorithms, which are used for system identification, require a restrictive condition of persistence of excitation (PE) for the estimates to converge to the true parameters. Some recent works have shown that collaborative system identification using multiple agents can relax the PE condition to a milder condition of collective persistence of excitation (C-PE) for parameter convergence. The C-PE condition implies that the PE condition is cooperatively satisfied by all the agents through sharing information between neighbors using a connected graph architecture, where each individual agent does not require to satisfy the PE condition separately. The proposed work designs a novel collaborative parameter estimator dynamics, which with the help of integral-like component ensures parameter convergence under a further slackened condition; coined as collective Initial Excitation (C-IE). The C-IE condition is an extension of the concept of initial excitation (IE), which is recently proposed in the context of parameter estimation in adaptive control. It has been already established that IE condition is significantly less restrictive than PE. The current work generalizes the concept of IE in a multi-agent settings, where information sharing through connected graph guarantees consensus parameter convergence under the C-IE condition. It can be argued that C-IE condition is milder than all of the other above mentioned conditions of PE, C-PE and IE. Simulation results further validate the efficacy of the proposed estimation algorithm.","PeriodicalId":6729,"journal":{"name":"2019 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"4 1","pages":"1285-1291"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77643044","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 : 2019-12-01DOI: 10.1109/SSCI44817.2019.9003089
Tony Jan, Saeid Iranmanesh, A. Sajeev
This paper proposes an innovative machine learning algorithm for resource optimization in IoT fog network. The proposed model utilizes distributed semi-supervised learning with innovative ensemble learning for efficient resource optimization in the IoT fog network for improved availability and readiness. The proposed model shows a great potential for real-time IoT applications utilizing the efficient fog resource optimization. The proposed model is evaluated against other state-of the-art models using the benchmark data to demonstrate its readiness and usefulness in real-time mission critical IoT applications such as in unmanned vehicle control system. The proposed model shows an acceptable resource optimization performance with reasonable computational complexity which proves to be useful in real-time IoT applications.
{"title":"Ensemble of Semi-Parametric Models for IoT Fog Modeling","authors":"Tony Jan, Saeid Iranmanesh, A. Sajeev","doi":"10.1109/SSCI44817.2019.9003089","DOIUrl":"https://doi.org/10.1109/SSCI44817.2019.9003089","url":null,"abstract":"This paper proposes an innovative machine learning algorithm for resource optimization in IoT fog network. The proposed model utilizes distributed semi-supervised learning with innovative ensemble learning for efficient resource optimization in the IoT fog network for improved availability and readiness. The proposed model shows a great potential for real-time IoT applications utilizing the efficient fog resource optimization. The proposed model is evaluated against other state-of the-art models using the benchmark data to demonstrate its readiness and usefulness in real-time mission critical IoT applications such as in unmanned vehicle control system. The proposed model shows an acceptable resource optimization performance with reasonable computational complexity which proves to be useful in real-time IoT applications.","PeriodicalId":6729,"journal":{"name":"2019 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"10 1","pages":"2995-2998"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"91516445","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 : 2019-12-01DOI: 10.1109/SSCI44817.2019.9002948
Thy Nguyen, Tayo Obafemi-Ajayi
In many applications, data exists in a mixed data type format, i.e. a combination of nominal (categorical) and numericalal features. A common practice for working with categorical features is to use an encoding method to transform the discrete values into numeric representation. However, numeric representation often neglects the innate structures in categorical features, potentially degrading the performance of learning algorithms. Utilizing the numeric representation could also limit interpretation of the learned model, such as finding the most discriminative categorical features or filtering irrelevant attributes. In this work, we extend the iterative hard thresholding (IHT) algorithm to quantify the structure of categorical features. The empirical evaluation of the proposed structured hard thresholding algorithm is based on both real and synthetic data sets in comparison with the original hard thresholding algorithm, LASSO and Random Forest. The results demonstrate an improved performance over the original IHT.
{"title":"Structured Iterative Hard Thresholding for Categorical and Mixed Data Types","authors":"Thy Nguyen, Tayo Obafemi-Ajayi","doi":"10.1109/SSCI44817.2019.9002948","DOIUrl":"https://doi.org/10.1109/SSCI44817.2019.9002948","url":null,"abstract":"In many applications, data exists in a mixed data type format, i.e. a combination of nominal (categorical) and numericalal features. A common practice for working with categorical features is to use an encoding method to transform the discrete values into numeric representation. However, numeric representation often neglects the innate structures in categorical features, potentially degrading the performance of learning algorithms. Utilizing the numeric representation could also limit interpretation of the learned model, such as finding the most discriminative categorical features or filtering irrelevant attributes. In this work, we extend the iterative hard thresholding (IHT) algorithm to quantify the structure of categorical features. The empirical evaluation of the proposed structured hard thresholding algorithm is based on both real and synthetic data sets in comparison with the original hard thresholding algorithm, LASSO and Random Forest. The results demonstrate an improved performance over the original IHT.","PeriodicalId":6729,"journal":{"name":"2019 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"19 1","pages":"2541-2547"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"91280633","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 : 2019-12-01DOI: 10.1109/SSCI44817.2019.9003124
Leo Ghignone, M. Barlow
Neural Networks are commonly used as function approximators in Reinforcement Learning, and the Extreme Learning Machine is one of the best algorithms to quickly train a shallow network. The online and sequential version OS-ELM could be a great candidate to quickly train a network to be a function approximator for Reinforcement Learning, but due to its non-forgetting properties it is actually not suitable for direct use with value estimations that improve in accuracy over time. This paper presents an alternative Neural Network training algorithm based on OS-ELM, which is able to perform learning online while dynamically modifying the weights of previously learned samples in order to decrease the importance of old samples learned over time. A mathematical derivation of the formulas used is presented, along with results of experiments showing equivalence of our algorithm to ELM when learning classic datasets and the advantage provided when dealing with Reinforcement Learning data.
{"title":"Shallow Network Training With Dynamic Sample Weights Decay - a Potential Function Approximator for Reinforcement Learning","authors":"Leo Ghignone, M. Barlow","doi":"10.1109/SSCI44817.2019.9003124","DOIUrl":"https://doi.org/10.1109/SSCI44817.2019.9003124","url":null,"abstract":"Neural Networks are commonly used as function approximators in Reinforcement Learning, and the Extreme Learning Machine is one of the best algorithms to quickly train a shallow network. The online and sequential version OS-ELM could be a great candidate to quickly train a network to be a function approximator for Reinforcement Learning, but due to its non-forgetting properties it is actually not suitable for direct use with value estimations that improve in accuracy over time. This paper presents an alternative Neural Network training algorithm based on OS-ELM, which is able to perform learning online while dynamically modifying the weights of previously learned samples in order to decrease the importance of old samples learned over time. A mathematical derivation of the formulas used is presented, along with results of experiments showing equivalence of our algorithm to ELM when learning classic datasets and the advantage provided when dealing with Reinforcement Learning data.","PeriodicalId":6729,"journal":{"name":"2019 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"11 1","pages":"149-154"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90294998","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 : 2019-12-01DOI: 10.1109/SSCI44817.2019.9003036
Huichao Zhang, Guangyuan Zhao, X. Qin
Scientific grain storage plays an important role in ensuring food security and promoting high-efficiency energy-saving operations. The paper provides more accurate reference datas for grain storage work. It can easily monitor the grain situation during the reserve period, and can scientifically predict the future grain development trend more accurately. It takes countermeasure in advance to prevent food disaster and further reduce. The workload of the warehouse clerk and related staff, while ensuring the safe and stable operation of the grain storage. Compared with the traditional Gery Model, the residual correction method is proposed to improve the data prediction accuracy. Combined with the Grey Verhulst model, a new residual-corrected Verhulst model is proposed. The simulation prove that the improved model is more traditional than the traditional one. The model is more conducive to the prediction of volatility data and the prediction accuracy is greatly improved.
{"title":"Analysis of Grain Condition in Improved Granary Based on Grey Prediction Algorithm","authors":"Huichao Zhang, Guangyuan Zhao, X. Qin","doi":"10.1109/SSCI44817.2019.9003036","DOIUrl":"https://doi.org/10.1109/SSCI44817.2019.9003036","url":null,"abstract":"Scientific grain storage plays an important role in ensuring food security and promoting high-efficiency energy-saving operations. The paper provides more accurate reference datas for grain storage work. It can easily monitor the grain situation during the reserve period, and can scientifically predict the future grain development trend more accurately. It takes countermeasure in advance to prevent food disaster and further reduce. The workload of the warehouse clerk and related staff, while ensuring the safe and stable operation of the grain storage. Compared with the traditional Gery Model, the residual correction method is proposed to improve the data prediction accuracy. Combined with the Grey Verhulst model, a new residual-corrected Verhulst model is proposed. The simulation prove that the improved model is more traditional than the traditional one. The model is more conducive to the prediction of volatility data and the prediction accuracy is greatly improved.","PeriodicalId":6729,"journal":{"name":"2019 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"64 1","pages":"2926-2932"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90572018","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 : 2019-12-01DOI: 10.1109/SSCI44817.2019.9002822
A. Zonta, S. Smit, M. Hoogendoorn, A. Eiben
Modelling human behaviour in simulation is still an ongoing challenge that spaces between several fields like social science, artificial intelligence, and philosophy. Humans normally move driven by their intent (e.g. to get groceries) and the surrounding environment (e.g. curiosity to see new interesting places). Normal services available online and offline do not consider the environment when planning the path. Especially on a leisure trip, this is very important. This paper presents a comparison between different machine learning algorithms and a famous path planning algorithm in the task of generating human-like trajectories based on environmental features. We show how a modified version of the well known A* algorithm outperforms different machine learning algorithms by computed evaluation metrics and human evaluation in the task of generating bike trips in the area around Ljubljana, Slovenia.
{"title":"Generation of Human-Like Movements Based on Environmental Features","authors":"A. Zonta, S. Smit, M. Hoogendoorn, A. Eiben","doi":"10.1109/SSCI44817.2019.9002822","DOIUrl":"https://doi.org/10.1109/SSCI44817.2019.9002822","url":null,"abstract":"Modelling human behaviour in simulation is still an ongoing challenge that spaces between several fields like social science, artificial intelligence, and philosophy. Humans normally move driven by their intent (e.g. to get groceries) and the surrounding environment (e.g. curiosity to see new interesting places). Normal services available online and offline do not consider the environment when planning the path. Especially on a leisure trip, this is very important. This paper presents a comparison between different machine learning algorithms and a famous path planning algorithm in the task of generating human-like trajectories based on environmental features. We show how a modified version of the well known A* algorithm outperforms different machine learning algorithms by computed evaluation metrics and human evaluation in the task of generating bike trips in the area around Ljubljana, Slovenia.","PeriodicalId":6729,"journal":{"name":"2019 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"50 1","pages":"3079-3086"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85667583","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 : 2019-12-01DOI: 10.1109/SSCI44817.2019.9002689
Liyu Lin, Shuanqiang Yang
According to the shortcomings of the traditional two-dimensional Otsu threshold method in segmentation accuracy and anti-production performance, an improved method based on weighted two-dimensional Otsu threshold segmentation image is proposed. On the basis of the cross-division of two-dimensional histogram, the distribution information of gray level and probability of gray value is used to comprehensively consider the influence of inter-class variance and intra-class variance on image segmentation, and the threshold value is weighted by the ratio of target and background in the image, which makes the threshold value closer to the ideal segmentation threshold. Finally, the simulation experiment is carried out to verify that the improved weighted segmentation method can achieve a good image segmentation effect and have better anti-noise ability.
{"title":"Weighted Two-dimensional Otsu Threshold Approached for Image Segmentation","authors":"Liyu Lin, Shuanqiang Yang","doi":"10.1109/SSCI44817.2019.9002689","DOIUrl":"https://doi.org/10.1109/SSCI44817.2019.9002689","url":null,"abstract":"According to the shortcomings of the traditional two-dimensional Otsu threshold method in segmentation accuracy and anti-production performance, an improved method based on weighted two-dimensional Otsu threshold segmentation image is proposed. On the basis of the cross-division of two-dimensional histogram, the distribution information of gray level and probability of gray value is used to comprehensively consider the influence of inter-class variance and intra-class variance on image segmentation, and the threshold value is weighted by the ratio of target and background in the image, which makes the threshold value closer to the ideal segmentation threshold. Finally, the simulation experiment is carried out to verify that the improved weighted segmentation method can achieve a good image segmentation effect and have better anti-noise ability.","PeriodicalId":6729,"journal":{"name":"2019 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"66 1","pages":"1002-1006"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88576153","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 : 2019-12-01DOI: 10.1109/SSCI44817.2019.9002917
Shumpei Matsuoka, Yao Jiang, A. Sasou
Fundamental frequency (F0) contours play a very important role in reflecting the emotion, identity, intension, and attitude of a speaker in samples of speech. In this paper, we adopted a generative adversarial network (GAN) to generate artificial F0 contours of emotional speech. The GAN faces some limitations, however, in that it frequently generates undesired data because of unstable training, and it can repeatedly generate very similar or the same data, which is known as mode collapse. This study constructed a GAN-based generative model for F0 contours that can stably generate more-various F0 contours that fit the statistical characteristics of the training data. We tested the classification rate of four kinds of emotions in the F0 contours generated from five kinds of generative models. We also evaluated the averaged local density of the generated F0 contours to represent the variety of the generated F0 contours. Preliminary experiments confirmed the validity and effectiveness of the proposed generative model.
{"title":"Generation of Artificial FO-contours of Emotional Speech with Generative Adversarial Networks","authors":"Shumpei Matsuoka, Yao Jiang, A. Sasou","doi":"10.1109/SSCI44817.2019.9002917","DOIUrl":"https://doi.org/10.1109/SSCI44817.2019.9002917","url":null,"abstract":"Fundamental frequency (F0) contours play a very important role in reflecting the emotion, identity, intension, and attitude of a speaker in samples of speech. In this paper, we adopted a generative adversarial network (GAN) to generate artificial F0 contours of emotional speech. The GAN faces some limitations, however, in that it frequently generates undesired data because of unstable training, and it can repeatedly generate very similar or the same data, which is known as mode collapse. This study constructed a GAN-based generative model for F0 contours that can stably generate more-various F0 contours that fit the statistical characteristics of the training data. We tested the classification rate of four kinds of emotions in the F0 contours generated from five kinds of generative models. We also evaluated the averaged local density of the generated F0 contours to represent the variety of the generated F0 contours. Preliminary experiments confirmed the validity and effectiveness of the proposed generative model.","PeriodicalId":6729,"journal":{"name":"2019 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"27 1","pages":"1030-1034"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88729388","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 : 2019-12-01DOI: 10.1109/SSCI44817.2019.9003133
Max Schüssler, T. Münker, O. Nelles
Deep recurrent neural networks are used as a means for nonlinear system identification. It is shown that state space models can be transformed into recurrent neural networks and vice versa. This transformation and the understanding of the long short-term memory cell in terms of common system identification nomenclature makes the advances in deep learning more accessible to the controls and system identification communities. A systematic study of deep recurrent neural networks is carried out on a state-of-the-art system identification benchmark. The results indicate that if high amounts of data are available, standard recurrent neural networks achieve comparable performance to state-of-the-art system identification methods.
{"title":"Deep Recurrent Neural Networks for Nonlinear System Identification","authors":"Max Schüssler, T. Münker, O. Nelles","doi":"10.1109/SSCI44817.2019.9003133","DOIUrl":"https://doi.org/10.1109/SSCI44817.2019.9003133","url":null,"abstract":"Deep recurrent neural networks are used as a means for nonlinear system identification. It is shown that state space models can be transformed into recurrent neural networks and vice versa. This transformation and the understanding of the long short-term memory cell in terms of common system identification nomenclature makes the advances in deep learning more accessible to the controls and system identification communities. A systematic study of deep recurrent neural networks is carried out on a state-of-the-art system identification benchmark. The results indicate that if high amounts of data are available, standard recurrent neural networks achieve comparable performance to state-of-the-art system identification methods.","PeriodicalId":6729,"journal":{"name":"2019 IEEE Symposium Series on Computational Intelligence (SSCI)","volume":"1 1","pages":"448-454"},"PeriodicalIF":0.0,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88886411","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：