Pub Date : 2013-08-01DOI: 10.1109/IJCNN.2013.6707139
Huang Chao-hui
The vertebral pose is critical information in orthopedics. An automated vertebral pose estimation can provide direct supports to medical diagnoses. In this paper, we proposed a vertebral pose estimation based on the given two sets of training patterns. The first set contains the images of vertebrae, in which all vertebral columns are fixed at a proper pose; the second are the images which are cropped with arbitrarily shift and rotation. Based on these two pattern sets, the proposed method can perform template matching. By using exhaustive searching, we will be able to estimate the poses of the vertebral columns on the given x-ray images. We propose a new approach for extracting critical information from the given training patterns in the problems of classification. In this work, we use it to estimate the poses of vertebral columns on x-ray images. The proposed method consists of two parts: 1, feature extraction and 2, classification. the first part extracts the unique features from the two given training pattern sets. These unique features are used to support the second part, which is a classifier inspired by the famous AdaBoost.
{"title":"Pose estimation for vertebral mobility analysis using eXclusive-ICA based boosting (XICABoost) algorithm","authors":"Huang Chao-hui","doi":"10.1109/IJCNN.2013.6707139","DOIUrl":"https://doi.org/10.1109/IJCNN.2013.6707139","url":null,"abstract":"The vertebral pose is critical information in orthopedics. An automated vertebral pose estimation can provide direct supports to medical diagnoses. In this paper, we proposed a vertebral pose estimation based on the given two sets of training patterns. The first set contains the images of vertebrae, in which all vertebral columns are fixed at a proper pose; the second are the images which are cropped with arbitrarily shift and rotation. Based on these two pattern sets, the proposed method can perform template matching. By using exhaustive searching, we will be able to estimate the poses of the vertebral columns on the given x-ray images. We propose a new approach for extracting critical information from the given training patterns in the problems of classification. In this work, we use it to estimate the poses of vertebral columns on x-ray images. The proposed method consists of two parts: 1, feature extraction and 2, classification. the first part extracts the unique features from the two given training pattern sets. These unique features are used to support the second part, which is a classifier inspired by the famous AdaBoost.","PeriodicalId":376975,"journal":{"name":"The 2013 International Joint Conference on Neural Networks (IJCNN)","volume":"128 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121026762","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 : 2013-08-01DOI: 10.1109/IJCNN.2013.6706812
Gangyi Wang, Guanghui Ren, Zhilu Wu, Yaqin Zhao, Lihui Jiang
We present a traffic sign detection method which has won the first place for the prohibitory and mandatory signs and the third place for the danger signs in the GTSDB competition. The method uses the histogram of oriented gradient (HOG) and a coarse-to-fine sliding window scheme. Candidate ROIs are first roughly detected within a small-sized window, and then further verified within a large-sized window for higher accuracy. Experimental results show that the proposed method achieves high recall and precision ratios, and is robust to various adverse situations including bad lighting condition, partial occlusion, low quality and small projective deformation.
{"title":"A robust, coarse-to-fine traffic sign detection method","authors":"Gangyi Wang, Guanghui Ren, Zhilu Wu, Yaqin Zhao, Lihui Jiang","doi":"10.1109/IJCNN.2013.6706812","DOIUrl":"https://doi.org/10.1109/IJCNN.2013.6706812","url":null,"abstract":"We present a traffic sign detection method which has won the first place for the prohibitory and mandatory signs and the third place for the danger signs in the GTSDB competition. The method uses the histogram of oriented gradient (HOG) and a coarse-to-fine sliding window scheme. Candidate ROIs are first roughly detected within a small-sized window, and then further verified within a large-sized window for higher accuracy. Experimental results show that the proposed method achieves high recall and precision ratios, and is robust to various adverse situations including bad lighting condition, partial occlusion, low quality and small projective deformation.","PeriodicalId":376975,"journal":{"name":"The 2013 International Joint Conference on Neural Networks (IJCNN)","volume":"68 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121054979","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 : 2013-08-01DOI: 10.1109/IJCNN.2013.6707055
Xiaofeng Lin, N. Cao, Yuzhang Lin
In this paper, a new finite horizon iterative ADP algorithm is used to solve a class of nonlinear systems with state delay and control constraints problem and finite time ε-optimal control is obtained. First of all, a new performance index function is designed to deal with the control constraints, the discrete nonlinear systems HJB equation with state delay is presented. Second, the iterative process and mathematical proof of the convergence is illustrated for the proposed finite horizon ADP algorithm. Approximate optimal control is obtained by introducing an error bond ε. Two BP neural networks are developed to approximate control law function and performance index function in our ADP algorithm. Finally, comparing simulation cases are used to verify the effectiveness of the method proposed in this paper.
{"title":"Neural network based finite horizon optimal control for a class of nonlinear systems with state delay and control constraints","authors":"Xiaofeng Lin, N. Cao, Yuzhang Lin","doi":"10.1109/IJCNN.2013.6707055","DOIUrl":"https://doi.org/10.1109/IJCNN.2013.6707055","url":null,"abstract":"In this paper, a new finite horizon iterative ADP algorithm is used to solve a class of nonlinear systems with state delay and control constraints problem and finite time ε-optimal control is obtained. First of all, a new performance index function is designed to deal with the control constraints, the discrete nonlinear systems HJB equation with state delay is presented. Second, the iterative process and mathematical proof of the convergence is illustrated for the proposed finite horizon ADP algorithm. Approximate optimal control is obtained by introducing an error bond ε. Two BP neural networks are developed to approximate control law function and performance index function in our ADP algorithm. Finally, comparing simulation cases are used to verify the effectiveness of the method proposed in this paper.","PeriodicalId":376975,"journal":{"name":"The 2013 International Joint Conference on Neural Networks (IJCNN)","volume":"235 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121246924","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 : 2013-08-01DOI: 10.1109/IJCNN.2013.6706967
Mojtaba Sedigh Fazli, Jean-Fabrice Lebraty
Forecasting in a risky situation is a very important function for managers to assist them in decision-making. One of the fluctuated markets in stock exchange market is chemical market. In this research the target item for prediction is PET (Poly Ethylene Terephthalate) which is the raw material for textile industries and it's very sensitive on oil prices and the demand and supply ratio. The main idea is coming through NORN model which was presented by Lee and Liu [1]. In this article after modifying the NORN model, a model has been proposed and real data are applied to this new model (we named it AHIS which stands for Adaptive Hybrid Intelligent System). Finally, three different types of simulation have been conducted and compared with each other. They show that hybrid model which is supporting both Fuzzy Systems and Neural Networks concepts, satisfied the research question considerably. In normal situation the model forecasts a relevant trend and can be used as a DSS for a manager.
{"title":"A comparative study on forecasting polyester chips prices for 15 days, using different hybrid intelligent systems","authors":"Mojtaba Sedigh Fazli, Jean-Fabrice Lebraty","doi":"10.1109/IJCNN.2013.6706967","DOIUrl":"https://doi.org/10.1109/IJCNN.2013.6706967","url":null,"abstract":"Forecasting in a risky situation is a very important function for managers to assist them in decision-making. One of the fluctuated markets in stock exchange market is chemical market. In this research the target item for prediction is PET (Poly Ethylene Terephthalate) which is the raw material for textile industries and it's very sensitive on oil prices and the demand and supply ratio. The main idea is coming through NORN model which was presented by Lee and Liu [1]. In this article after modifying the NORN model, a model has been proposed and real data are applied to this new model (we named it AHIS which stands for Adaptive Hybrid Intelligent System). Finally, three different types of simulation have been conducted and compared with each other. They show that hybrid model which is supporting both Fuzzy Systems and Neural Networks concepts, satisfied the research question considerably. In normal situation the model forecasts a relevant trend and can be used as a DSS for a manager.","PeriodicalId":376975,"journal":{"name":"The 2013 International Joint Conference on Neural Networks (IJCNN)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123693777","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 : 2013-08-01DOI: 10.1109/IJCNN.2013.6707041
A. Rizzi, Francesca Possemato, L. Livi, Azzurra Sebastiani, A. Giuliani, F. Mascioli
In this paper we propose a classifier for generalized sequences that is conceived in the granular computing framework. The classification system processes the input sequences of objects by means of a suited interplay among dissimilarity and clustering based techniques. The core data mining engine retrieves information granules that are used to represent the input sequences as feature vectors. Such a representation allows to deal with the original sequence classification problem through standard pattern recognition tools. We have evaluated the generalization capability of the system in an interesting case study concerning the protein folding problem. In the considered dataset, the entire E. Coli proteome was screened as for the prediction of protein relative solubility on a pure amino acids sequence basis. We report the analysis of the dataset considering different settings, showing interesting test set classification accuracy results. The developed system consents also to extract knowledge from the considered training set, by allowing the analysis of the retrieved information granules.
{"title":"A dissimilarity-based classifier for generalized sequences by a granular computing approach","authors":"A. Rizzi, Francesca Possemato, L. Livi, Azzurra Sebastiani, A. Giuliani, F. Mascioli","doi":"10.1109/IJCNN.2013.6707041","DOIUrl":"https://doi.org/10.1109/IJCNN.2013.6707041","url":null,"abstract":"In this paper we propose a classifier for generalized sequences that is conceived in the granular computing framework. The classification system processes the input sequences of objects by means of a suited interplay among dissimilarity and clustering based techniques. The core data mining engine retrieves information granules that are used to represent the input sequences as feature vectors. Such a representation allows to deal with the original sequence classification problem through standard pattern recognition tools. We have evaluated the generalization capability of the system in an interesting case study concerning the protein folding problem. In the considered dataset, the entire E. Coli proteome was screened as for the prediction of protein relative solubility on a pure amino acids sequence basis. We report the analysis of the dataset considering different settings, showing interesting test set classification accuracy results. The developed system consents also to extract knowledge from the considered training set, by allowing the analysis of the retrieved information granules.","PeriodicalId":376975,"journal":{"name":"The 2013 International Joint Conference on Neural Networks (IJCNN)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125513394","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 : 2013-08-01DOI: 10.1109/IJCNN.2013.6706890
Guoqiang Xu, S. Furao, Jinxi Zhao
This paper studies empirically the effect of different sampling methods on training classifiers on the imbalanced data of the BCI P300 Speller. Both over-sampling and under-sampling are considered. Besides some existing methods like SMOTE that have been shown to be effective in addressing the class imbalance problem we also proposed a new under-sampling technology, namely, instance-remove algorithm which is based on the property of P300 data sets. The classifiers for testing are FLDA and linear SVM. Experimental results suggest that not all of the sampling methods are effective in P300 detection, and even the same method may have different influence on different classifiers. It reveals that the SMOTE technique which is a variant of over-sampling is very effective in training an FLDA classifier while other methods are slightly effective or ineffective both in training FLDA and Linear SVM. The study also suggests that the over-sampling is more effective than under-sampling on both classifiers.
{"title":"The effect of methods addressing the class imbalance problem on P300 detection","authors":"Guoqiang Xu, S. Furao, Jinxi Zhao","doi":"10.1109/IJCNN.2013.6706890","DOIUrl":"https://doi.org/10.1109/IJCNN.2013.6706890","url":null,"abstract":"This paper studies empirically the effect of different sampling methods on training classifiers on the imbalanced data of the BCI P300 Speller. Both over-sampling and under-sampling are considered. Besides some existing methods like SMOTE that have been shown to be effective in addressing the class imbalance problem we also proposed a new under-sampling technology, namely, instance-remove algorithm which is based on the property of P300 data sets. The classifiers for testing are FLDA and linear SVM. Experimental results suggest that not all of the sampling methods are effective in P300 detection, and even the same method may have different influence on different classifiers. It reveals that the SMOTE technique which is a variant of over-sampling is very effective in training an FLDA classifier while other methods are slightly effective or ineffective both in training FLDA and Linear SVM. The study also suggests that the over-sampling is more effective than under-sampling on both classifiers.","PeriodicalId":376975,"journal":{"name":"The 2013 International Joint Conference on Neural Networks (IJCNN)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125693888","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 : 2013-08-01DOI: 10.1109/IJCNN.2013.6706944
M. G. Quiles, E. Zorzal, E. Macau
One important feature observed in several complex networks is the structure of communities, or modular structure. Detecting communities is still a big challenge for researchers, specially the development of models to deal with dynamic networks. Here, we propose a new method for detecting communities by using a dynamical model. The first step consists of generating a spatial representation, named particle, for each vertex in the network. With these two representation, network structure and the spatial particles, we define the model's dynamics by means of two interactions types: the first is related to the network structure, or relational, and it is responsible for approaching particles representing neighbor vertices; the second, repulsive, is generated according to the spatial position of each particle and is responsible to make each unrelated particle, according to the network structure, to repel each other. Thus, after a couple of iteration, we observe the formation of groups of particles representing communities. On the other hand, distinct communities are separated according to the spatial positions of their particles. Simulation results show that our model achieves good results on the two benchmark models taken into account and that it can also deal with dynamic networks owing to its intrinsic dynamics.
{"title":"A dynamical model for community detection in complex networks","authors":"M. G. Quiles, E. Zorzal, E. Macau","doi":"10.1109/IJCNN.2013.6706944","DOIUrl":"https://doi.org/10.1109/IJCNN.2013.6706944","url":null,"abstract":"One important feature observed in several complex networks is the structure of communities, or modular structure. Detecting communities is still a big challenge for researchers, specially the development of models to deal with dynamic networks. Here, we propose a new method for detecting communities by using a dynamical model. The first step consists of generating a spatial representation, named particle, for each vertex in the network. With these two representation, network structure and the spatial particles, we define the model's dynamics by means of two interactions types: the first is related to the network structure, or relational, and it is responsible for approaching particles representing neighbor vertices; the second, repulsive, is generated according to the spatial position of each particle and is responsible to make each unrelated particle, according to the network structure, to repel each other. Thus, after a couple of iteration, we observe the formation of groups of particles representing communities. On the other hand, distinct communities are separated according to the spatial positions of their particles. Simulation results show that our model achieves good results on the two benchmark models taken into account and that it can also deal with dynamic networks owing to its intrinsic dynamics.","PeriodicalId":376975,"journal":{"name":"The 2013 International Joint Conference on Neural Networks (IJCNN)","volume":"123 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127049245","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 : 2013-08-01DOI: 10.1109/IJCNN.2013.6707123
S. Pal, Alireza Alaei, U. Pal, M. Blumenstein
Among all of the biometric authentication systems, handwritten signatures are considered as the most legally and socially accepted attributes for personal verification. The objective of this paper is to present an empirical contribution towards the understanding of a threshold-based signature verification technique involving off-line Bangla (Bengali) signatures. Experiments on signature verification involving non-English signatures are an important consideration in the signature verification area. Only very few research works employing signatures of Indian script have been considered in the field of non-English signature verification. To fill this gap, a threshold-based scheme for verification considering off-line Bangla signatures is proposed. Some techniques such as under-sampled bitmap, intersection/endpoint and directional chain code are employed for feature extraction. The Nearest Neighbour method is considered for classification. Furthermore, a Bangla signature database, which consists of 2400 (100×24) genuine signatures and 3000 (100×30) forgeries has been created and is employed for experimentation. We obtained a 15.57% Average Error Rate (AER) as the best verification result using directional chain code features employed in this research work.
{"title":"Off-line Bangla signature verification: An empirical study","authors":"S. Pal, Alireza Alaei, U. Pal, M. Blumenstein","doi":"10.1109/IJCNN.2013.6707123","DOIUrl":"https://doi.org/10.1109/IJCNN.2013.6707123","url":null,"abstract":"Among all of the biometric authentication systems, handwritten signatures are considered as the most legally and socially accepted attributes for personal verification. The objective of this paper is to present an empirical contribution towards the understanding of a threshold-based signature verification technique involving off-line Bangla (Bengali) signatures. Experiments on signature verification involving non-English signatures are an important consideration in the signature verification area. Only very few research works employing signatures of Indian script have been considered in the field of non-English signature verification. To fill this gap, a threshold-based scheme for verification considering off-line Bangla signatures is proposed. Some techniques such as under-sampled bitmap, intersection/endpoint and directional chain code are employed for feature extraction. The Nearest Neighbour method is considered for classification. Furthermore, a Bangla signature database, which consists of 2400 (100×24) genuine signatures and 3000 (100×30) forgeries has been created and is employed for experimentation. We obtained a 15.57% Average Error Rate (AER) as the best verification result using directional chain code features employed in this research work.","PeriodicalId":376975,"journal":{"name":"The 2013 International Joint Conference on Neural Networks (IJCNN)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116037008","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 : 2013-08-01DOI: 10.1109/IJCNN.2013.6707116
H. El-Bakry, Mohamed Hamada
In this paper, a new fast neural model for testing massive volume of medical data is presented. The idea is to accelerate the process of detecting and classifying pediatric respiratory diseases by using neural networks. This is done by applying cross correlation between the input patterns and the input weights of neural networks in the frequency domain rather than time domain. Furthermore, such model is very useful for understanding the internal relation between the medical patterns. In addition, the input patterns are collected in one vector and manipulated as a one pattern. Moreover, before training neural networks, rough sets are used to reduce the length of the feature input vector. The most important feature elements are used to train the neural networks. The reduced input medical patterns are classified to one of eight diseases. Simulation results confirm the theoretical considerations as 98% of all tested cases are classified correctly. The presented model can be applied successfully for any other classification application.
{"title":"Fast diagnosing of pediatric respiratory diseases by using high speed neural networks","authors":"H. El-Bakry, Mohamed Hamada","doi":"10.1109/IJCNN.2013.6707116","DOIUrl":"https://doi.org/10.1109/IJCNN.2013.6707116","url":null,"abstract":"In this paper, a new fast neural model for testing massive volume of medical data is presented. The idea is to accelerate the process of detecting and classifying pediatric respiratory diseases by using neural networks. This is done by applying cross correlation between the input patterns and the input weights of neural networks in the frequency domain rather than time domain. Furthermore, such model is very useful for understanding the internal relation between the medical patterns. In addition, the input patterns are collected in one vector and manipulated as a one pattern. Moreover, before training neural networks, rough sets are used to reduce the length of the feature input vector. The most important feature elements are used to train the neural networks. The reduced input medical patterns are classified to one of eight diseases. Simulation results confirm the theoretical considerations as 98% of all tested cases are classified correctly. The presented model can be applied successfully for any other classification application.","PeriodicalId":376975,"journal":{"name":"The 2013 International Joint Conference on Neural Networks (IJCNN)","volume":"97 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116104927","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 : 2013-08-01DOI: 10.1109/IJCNN.2013.6706997
Rohitash Chandra
Cooperative coevolution employs different problem decomposition methods to decompose the neural network problem into subcomponents. The efficiency of a problem decomposition method is dependent on the neural network architecture and the nature of the training problem. The adaptation of problem decomposition methods has been recently proposed which showed that different problem decomposition methods are needed at different phases in the evolutionary process. This paper employs an adaptive cooperative coevolution problem decomposition framework for training recurrent neural networks on chaotic time series problems. The Mackey Glass, Lorenz and Sunspot chaotic time series are used. The results show improvement in performance in most cases, however, there are some limitations when compared to cooperative coevolution and other methods from literature.
{"title":"Adaptive problem decomposition in cooperative coevolution of recurrent networks for time series prediction","authors":"Rohitash Chandra","doi":"10.1109/IJCNN.2013.6706997","DOIUrl":"https://doi.org/10.1109/IJCNN.2013.6706997","url":null,"abstract":"Cooperative coevolution employs different problem decomposition methods to decompose the neural network problem into subcomponents. The efficiency of a problem decomposition method is dependent on the neural network architecture and the nature of the training problem. The adaptation of problem decomposition methods has been recently proposed which showed that different problem decomposition methods are needed at different phases in the evolutionary process. This paper employs an adaptive cooperative coevolution problem decomposition framework for training recurrent neural networks on chaotic time series problems. The Mackey Glass, Lorenz and Sunspot chaotic time series are used. The results show improvement in performance in most cases, however, there are some limitations when compared to cooperative coevolution and other methods from literature.","PeriodicalId":376975,"journal":{"name":"The 2013 International Joint Conference on Neural Networks (IJCNN)","volume":"433 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2013-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116184996","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：