Pub Date : 2002-12-10DOI: 10.1109/NEUREL.2002.1057990
A. Vučković, D. Popović, V. Radivojevic
We describe a novel method for classifying alert vs. drowsy states from one-second long sequences of full spectrum EEG recordings. This method uses time series of inter-hemispeheric and intra-hemispheric cross spectral densities of full spectrum EEG as input to an artificial neural network (ANN) with two discrete outputs: drowsy and alert. The experimental data were collected from 17 subjects. Two experts in EEG interpretation visually inspected the data and provided the necessary expertise for the training of an ANN. After several experiments we selected the learning vector quantization (LVQ) as the most suitable neural network and used the data from 5 subjects for the training. Classification properties of LVQ were validated using the data recorded from the remaining 12 subjects, whose EEG recordings have not been used for the training of the ANN. The statistics were used as a measure of potential applicability of the LVQ: the t-distribution showed that in 95% (confidence interval) of the target group the matching between the human assessment and the network output was 94, 37/spl plusmn/1.95 percent.
{"title":"Artificial neural network for detecting drowsiness from EEG recordings","authors":"A. Vučković, D. Popović, V. Radivojevic","doi":"10.1109/NEUREL.2002.1057990","DOIUrl":"https://doi.org/10.1109/NEUREL.2002.1057990","url":null,"abstract":"We describe a novel method for classifying alert vs. drowsy states from one-second long sequences of full spectrum EEG recordings. This method uses time series of inter-hemispeheric and intra-hemispheric cross spectral densities of full spectrum EEG as input to an artificial neural network (ANN) with two discrete outputs: drowsy and alert. The experimental data were collected from 17 subjects. Two experts in EEG interpretation visually inspected the data and provided the necessary expertise for the training of an ANN. After several experiments we selected the learning vector quantization (LVQ) as the most suitable neural network and used the data from 5 subjects for the training. Classification properties of LVQ were validated using the data recorded from the remaining 12 subjects, whose EEG recordings have not been used for the training of the ANN. The statistics were used as a measure of potential applicability of the LVQ: the t-distribution showed that in 95% (confidence interval) of the target group the matching between the human assessment and the network output was 94, 37/spl plusmn/1.95 percent.","PeriodicalId":347066,"journal":{"name":"6th Seminar on Neural Network Applications in Electrical Engineering","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2002-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130037868","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 : 2002-12-10DOI: 10.1109/NEUREL.2002.1057972
M. Milosavljevic, L. Cander, S. Tomaśevič
In this work we further refine and improve the neural network based foF2 predictor, which is actually a neural autoregressive model with additional input signals (NNARX). Our analysis is focused on choice of X parts of NNARX model in order to capture middle and long term dependencies. Daily distribution of prediction error suggests need for structural changes of the neural network model, as well as adaptation of running average lengths used for determination of X inputs. Generalisation properties of proposed neural predictor are improved by carefully designed pruning procedure with additional regularisation term in criterion function.
{"title":"Ionospheric storm forecasting technique by artificial neural network","authors":"M. Milosavljevic, L. Cander, S. Tomaśevič","doi":"10.1109/NEUREL.2002.1057972","DOIUrl":"https://doi.org/10.1109/NEUREL.2002.1057972","url":null,"abstract":"In this work we further refine and improve the neural network based foF2 predictor, which is actually a neural autoregressive model with additional input signals (NNARX). Our analysis is focused on choice of X parts of NNARX model in order to capture middle and long term dependencies. Daily distribution of prediction error suggests need for structural changes of the neural network model, as well as adaptation of running average lengths used for determination of X inputs. Generalisation properties of proposed neural predictor are improved by carefully designed pruning procedure with additional regularisation term in criterion function.","PeriodicalId":347066,"journal":{"name":"6th Seminar on Neural Network Applications in Electrical Engineering","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2002-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125263726","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 : 2002-12-10DOI: 10.1109/NEUREL.2002.1057958
A. I. Hanna, I. Krcmar, D. Mandic
We attempt to explain in more detail the performance of several novel algorithms for nonlinear neural adaptive filtering. Weight trajectories together with the error surface give a clear understandable representation of the family of least mean square (LMS) based, nonlinear gradient descent (NGD), search-then-converge (STC) learning algorithms and the real-time recurrent learning (RTRL) algorithm. Performance is measured on prediction of coloured and nonlinear input. The results are an alternative qualitative representation of different qualitative performance measures for the analysed algorithms. Error surfaces and the adjacent instantaneous prediction errors support the analysis.
{"title":"Perlustration of error surfaces for nonlinear stochastic gradient descent algorithms","authors":"A. I. Hanna, I. Krcmar, D. Mandic","doi":"10.1109/NEUREL.2002.1057958","DOIUrl":"https://doi.org/10.1109/NEUREL.2002.1057958","url":null,"abstract":"We attempt to explain in more detail the performance of several novel algorithms for nonlinear neural adaptive filtering. Weight trajectories together with the error surface give a clear understandable representation of the family of least mean square (LMS) based, nonlinear gradient descent (NGD), search-then-converge (STC) learning algorithms and the real-time recurrent learning (RTRL) algorithm. Performance is measured on prediction of coloured and nonlinear input. The results are an alternative qualitative representation of different qualitative performance measures for the analysed algorithms. Error surfaces and the adjacent instantaneous prediction errors support the analysis.","PeriodicalId":347066,"journal":{"name":"6th Seminar on Neural Network Applications in Electrical Engineering","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2002-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130692398","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 : 2002-12-10DOI: 10.1109/NEUREL.2002.1057983
N. Stanic, M. Potrebić, D. Durdevic, D. Dujković, P. Kostic
An effective character recognition procedure is reported. The procedure uses a new architecture, that contains three blocks: a filter, a block with cellular neural network and a block for detection. An initial test result obtained shows 94-100% recognition rates for numerals.
{"title":"Character recognition using a cellular neural network","authors":"N. Stanic, M. Potrebić, D. Durdevic, D. Dujković, P. Kostic","doi":"10.1109/NEUREL.2002.1057983","DOIUrl":"https://doi.org/10.1109/NEUREL.2002.1057983","url":null,"abstract":"An effective character recognition procedure is reported. The procedure uses a new architecture, that contains three blocks: a filter, a block with cellular neural network and a block for detection. An initial test result obtained shows 94-100% recognition rates for numerals.","PeriodicalId":347066,"journal":{"name":"6th Seminar on Neural Network Applications in Electrical Engineering","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2002-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122929367","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 : 2002-12-10DOI: 10.1109/NEUREL.2002.1057986
R. Babic
Following a new paradigm of information encoding by spike timings and its processing by neurons as coincidence detectors, we first discuss some aspects of temporal neural phenomena, and give an evolutionary interpretation of the relationships between the axon diameter, propagation speed and density of neural tissue. Then we propose a recurrent architecture of neural network capable to convert periodic spike train into desired pattern of spike timings. Another configuration that we propose represent neural fiber as a delay element where the changeable delay effect is attained over lateral loops with creeping synapses which shortcut the spanned portions of the basic fiber. As the starting and termination might represent important indicators of a spike burst we also propose the structure of a neural differentiator with cross inhibition. Finally, we give the internal structure of a neural delay element with an incremental change of delay value, including an explanation of changing, i.e. the learning process.
{"title":"Some architectures of neural networks with temporal effects","authors":"R. Babic","doi":"10.1109/NEUREL.2002.1057986","DOIUrl":"https://doi.org/10.1109/NEUREL.2002.1057986","url":null,"abstract":"Following a new paradigm of information encoding by spike timings and its processing by neurons as coincidence detectors, we first discuss some aspects of temporal neural phenomena, and give an evolutionary interpretation of the relationships between the axon diameter, propagation speed and density of neural tissue. Then we propose a recurrent architecture of neural network capable to convert periodic spike train into desired pattern of spike timings. Another configuration that we propose represent neural fiber as a delay element where the changeable delay effect is attained over lateral loops with creeping synapses which shortcut the spanned portions of the basic fiber. As the starting and termination might represent important indicators of a spike burst we also propose the structure of a neural differentiator with cross inhibition. Finally, we give the internal structure of a neural delay element with an incremental change of delay value, including an explanation of changing, i.e. the learning process.","PeriodicalId":347066,"journal":{"name":"6th Seminar on Neural Network Applications in Electrical Engineering","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2002-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127333765","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 : 2002-12-10DOI: 10.1109/NEUREL.2002.1057975
Z. Durovic, B. Kovačević, V. Papic
The analysis of a class of neuro-fuzzy regulators, called ANFIS (adaptive neuro-fuzzy inherited systems), is presented in the paper. The overview of its structure and the training process is included. One example of nonlinear system is also used to illustrate the feasibility and the limits of this class of regulators.
{"title":"Analysis of one class of neuro-fuzzy regulators","authors":"Z. Durovic, B. Kovačević, V. Papic","doi":"10.1109/NEUREL.2002.1057975","DOIUrl":"https://doi.org/10.1109/NEUREL.2002.1057975","url":null,"abstract":"The analysis of a class of neuro-fuzzy regulators, called ANFIS (adaptive neuro-fuzzy inherited systems), is presented in the paper. The overview of its structure and the training process is included. One example of nonlinear system is also used to illustrate the feasibility and the limits of this class of regulators.","PeriodicalId":347066,"journal":{"name":"6th Seminar on Neural Network Applications in Electrical Engineering","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2002-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125921889","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 : 2002-12-10DOI: 10.1109/NEUREL.2002.1057968
P. Daras, D.K. Bechtsis, M. Strintzis
The objective of this paper is to present an experience in the challenge of using Internet and mobile Internet technologies for the development of a Web/WAP (wireless application protocol)-based medical application with data mining support. This medical application is focused on the development, diffusion and use of the technology in response to specific domain needs of medical experts in the area of cardiology, especially for the patients after by-pass operation.
{"title":"A Web/WAP-based system for remote monitoring patients with data mining support","authors":"P. Daras, D.K. Bechtsis, M. Strintzis","doi":"10.1109/NEUREL.2002.1057968","DOIUrl":"https://doi.org/10.1109/NEUREL.2002.1057968","url":null,"abstract":"The objective of this paper is to present an experience in the challenge of using Internet and mobile Internet technologies for the development of a Web/WAP (wireless application protocol)-based medical application with data mining support. This medical application is focused on the development, diffusion and use of the technology in response to specific domain needs of medical experts in the area of cardiology, especially for the patients after by-pass operation.","PeriodicalId":347066,"journal":{"name":"6th Seminar on Neural Network Applications in Electrical Engineering","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2002-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127873441","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 : 2002-12-10DOI: 10.1109/NEUREL.2002.1057997
A. Neskovic, N. Neskovic, D. Paunovic
A new microcell prediction model for mobile phone environment is presented in this paper. The model is based on the principles of popular feedforward neural networks. Utilising a new artificial neural network model some important disadvantages of both deterministic and empirical models can be overcome. In order to build the model, extensive electric field level measurements (in 900 MHz frequency band) were carried out in the city of Belgrade, for two different test transmitter locations. The comparison between the data obtained by the proposed electric field level prediction model and the independent measurement sets, have shown that the proposed model is accurate (on the order of the local mean measurements uncertainty) and reliable. At the same time, the algorithm is suitable for computer implementation, simple and fast.
{"title":"Microcell coverage prediction using artificial neural networks","authors":"A. Neskovic, N. Neskovic, D. Paunovic","doi":"10.1109/NEUREL.2002.1057997","DOIUrl":"https://doi.org/10.1109/NEUREL.2002.1057997","url":null,"abstract":"A new microcell prediction model for mobile phone environment is presented in this paper. The model is based on the principles of popular feedforward neural networks. Utilising a new artificial neural network model some important disadvantages of both deterministic and empirical models can be overcome. In order to build the model, extensive electric field level measurements (in 900 MHz frequency band) were carried out in the city of Belgrade, for two different test transmitter locations. The comparison between the data obtained by the proposed electric field level prediction model and the independent measurement sets, have shown that the proposed model is accurate (on the order of the local mean measurements uncertainty) and reliable. At the same time, the algorithm is suitable for computer implementation, simple and fast.","PeriodicalId":347066,"journal":{"name":"6th Seminar on Neural Network Applications in Electrical Engineering","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2002-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128020237","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 : 2002-12-10DOI: 10.1109/NEUREL.2002.1057955
P. Cristea
The complex representation of the nucleotides derived from the projection of the Nucleotide Tetrahedron on an adequately oriented plane is used to convert sequences of nucleotides into complex digital genomic signals. This approach offers the possibility to use signal processing methods for the handling and analyzing of genomic information at the nucleotide, codon and amino acid levels in a multiresolutional approach. Some basic features of nucleotide sequences can be elicited using these signal representations. Specifically, the paper presents large scale features of eukaryote and prokaryote DNA genomic signals obtained with phase analysis methods that reveal regularities in the statistics of base distribution and of base-to-base transitions distribution along the DNA strands.
{"title":"Whole chromosome features of genomic signals","authors":"P. Cristea","doi":"10.1109/NEUREL.2002.1057955","DOIUrl":"https://doi.org/10.1109/NEUREL.2002.1057955","url":null,"abstract":"The complex representation of the nucleotides derived from the projection of the Nucleotide Tetrahedron on an adequately oriented plane is used to convert sequences of nucleotides into complex digital genomic signals. This approach offers the possibility to use signal processing methods for the handling and analyzing of genomic information at the nucleotide, codon and amino acid levels in a multiresolutional approach. Some basic features of nucleotide sequences can be elicited using these signal representations. Specifically, the paper presents large scale features of eukaryote and prokaryote DNA genomic signals obtained with phase analysis methods that reveal regularities in the statistics of base distribution and of base-to-base transitions distribution along the DNA strands.","PeriodicalId":347066,"journal":{"name":"6th Seminar on Neural Network Applications in Electrical Engineering","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2002-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132724063","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 : 2002-12-10DOI: 10.1109/NEUREL.2002.1057999
S. I. Bank, Z. Avramovic
This paper considers the possibility of application Hopfield recurrent neural network in solving travelling salesman problem when nodes are given in sphere coordinates and when distances between nodes are not linear but sphere. Obtained numerical results in case of an arbitrary chosen example are presented.
{"title":"Hopfield network in solving travelling salesman problem in navigation","authors":"S. I. Bank, Z. Avramovic","doi":"10.1109/NEUREL.2002.1057999","DOIUrl":"https://doi.org/10.1109/NEUREL.2002.1057999","url":null,"abstract":"This paper considers the possibility of application Hopfield recurrent neural network in solving travelling salesman problem when nodes are given in sphere coordinates and when distances between nodes are not linear but sphere. Obtained numerical results in case of an arbitrary chosen example are presented.","PeriodicalId":347066,"journal":{"name":"6th Seminar on Neural Network Applications in Electrical Engineering","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2002-12-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127018976","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：