{"title":"ART-R: a novel reinforcement learning algorithm using an ART module for state representation","authors":"L. Brignone, M. Howarth","doi":"10.1109/NNSP.2003.1318082","DOIUrl":"https://doi.org/10.1109/NNSP.2003.1318082","url":null,"abstract":"The work introduces a neural network (NN) algorithm capable of merging the fast and stable learning behaviour offered by the adaptive resonance theory (ART) and the advantageous properties of a reinforcement learning agent. The result is ART-R a neural algorithm particularly suited to learning state-action mappings in control applications. A real time example addressing a typical problem found in autonomous robotic assembly is discussed to highlight the achievement of unsupervised and fast learning of an optimal behaviour.","PeriodicalId":315958,"journal":{"name":"2003 IEEE XIII Workshop on Neural Networks for Signal Processing (IEEE Cat. No.03TH8718)","volume":"82 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115600377","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}
{"title":"Neural speech enhancement in the time-frequency domain","authors":"M. Volkmer","doi":"10.1109/NNSP.2003.1318061","DOIUrl":"https://doi.org/10.1109/NNSP.2003.1318061","url":null,"abstract":"A pulsed neural network approach for the enhancement of speech is presented. It employs a recently introduced pulse-induced masking principle on a joint time-frequency representation of signals. Based on the notion of a spectro-temporal receptive field, the principle provides signal-dependent dictionary elements, which are integrated into an enhancement process in the time-frequency domain. Experiments on signals with different types of noise are conducted with the resulting neural time-frequency domain method.","PeriodicalId":315958,"journal":{"name":"2003 IEEE XIII Workshop on Neural Networks for Signal Processing (IEEE Cat. No.03TH8718)","volume":"108 2","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"120991657","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}
{"title":"A self-organizing method for map reconstruction","authors":"I. K. Altinel, N. Aras, B. Oommen","doi":"10.1109/NNSP.2003.1318067","DOIUrl":"https://doi.org/10.1109/NNSP.2003.1318067","url":null,"abstract":"A variety of problems in geographical and satellite-based remote sensing signal processing, and in the area of \"zero-error\" pattern recognition dealing with processing the information contained in the distances between the points in the geographical or feature space. In this paper we consider one such problem, namely, that of reconstructing the points in the geographical or feature space, when we are only given the approximate distances between the points themselves. In particular, we are interested in the problem of reconstructing a map when the given data is the set of intercity road travel distances. Reported solution approaches primarily involve multi-dimensional scaling techniques. However, we propose a self-organizing method. The new method is tested and compared with the classical multi-dimensional scaling and ALSCAL on different data sets obtained from various countries.","PeriodicalId":315958,"journal":{"name":"2003 IEEE XIII Workshop on Neural Networks for Signal Processing (IEEE Cat. No.03TH8718)","volume":"61 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116713189","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}
{"title":"Nonlinear adaptive prediction using a complex-valued PRNN","authors":"S. L. Goh, D. Mandic","doi":"10.1109/NNSP.2003.1318077","DOIUrl":"https://doi.org/10.1109/NNSP.2003.1318077","url":null,"abstract":"A computationally efficient architecture for nonlinear adaptive prediction of complex-valued nonlinear and non-stationary signals is presented. The adaptive predictor is based upon a complex-valued pipelined recurrent neural network (CPRNN) trained by the complex-valued real-time recurrent learning (CRTRL) algorithm. A variable forgetting factor (VFF) is introduced to improve the performance of CPRNN in the non-stationary environment. The analysis is undertaken with respect to the number of the nested modules, forgetting factor, and input memory of the CPRNN. Simulations on real and synthetic complex data support the proposed architecture and algorithms.","PeriodicalId":315958,"journal":{"name":"2003 IEEE XIII Workshop on Neural Networks for Signal Processing (IEEE Cat. No.03TH8718)","volume":"56 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122526215","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}
{"title":"Variogram based noise variance estimation and its use in kernel based regression","authors":"K. Pelckmans, J. Brabanter, J. Suykens, B. Moor","doi":"10.1109/NNSP.2003.1318019","DOIUrl":"https://doi.org/10.1109/NNSP.2003.1318019","url":null,"abstract":"Model-free estimates of the noise variance are important for doing model selection and setting tuning parameters. In this paper a data representation is discussed which leads to such an estimator suitable for multi-dimensional input data. The visual representation, called the differogram cloud, is based on the 2-norm of the differences of the input- and output-data. A corrected way to estimate the variance of the noise on the output measurement and a (tuning) parameter free version are derived. Connections with other existing variance estimators and numerical simulations indicate convergence of the estimators. As a special case, this paper focuses on model selection and tuning parameters of least squares support vector machines [J. Suykens, et al., 2002].","PeriodicalId":315958,"journal":{"name":"2003 IEEE XIII Workshop on Neural Networks for Signal Processing (IEEE Cat. No.03TH8718)","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124796488","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}
{"title":"On center locating in RBF network equalization in the GSM system","authors":"Arto Kantsila, M. Lehtokangas, J. Saarinen","doi":"10.1109/NNSP.2003.1318057","DOIUrl":"https://doi.org/10.1109/NNSP.2003.1318057","url":null,"abstract":"In this paper we have studied methods for center locating in radial basis function (RBF) network equalization in the GSM (Global System for Mobile Communications) environment. Here, equalization is considered as a classification problem, where the idea is to map the received complex-valued signal into desired binary values using RBF network equalizer. Two techniques for the RBF center locating have been studied. The first one applies a nearest-neighbor type clustering procedure and the second one uses estimated channel coefficients for computing the possible center locations. These methods are studied in terms of bit error rates and computational efficiency. Performance comparisons are made to a previously studied RBF network, which considers each received training sequence vector as a center and also to a Viterbi equalizer.","PeriodicalId":315958,"journal":{"name":"2003 IEEE XIII Workshop on Neural Networks for Signal Processing (IEEE Cat. No.03TH8718)","volume":"302 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133500261","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}
{"title":"A paradigm for blind IIR equalization using the constant modulus criterion and an artificial immune network","authors":"R. Attux, L. Castro, F. V. Zuben, J. Romano","doi":"10.1109/NNSP.2003.1318083","DOIUrl":"https://doi.org/10.1109/NNSP.2003.1318083","url":null,"abstract":"We propose a new paradigm for optimal blind IIR equalization using the well-known constant modulus criterion and an artificial immune network. Tests in three different scenarios reveal the efficiency of the proposal, attested by excellent global convergence rates (100% in two cases) and adaptation patterns.","PeriodicalId":315958,"journal":{"name":"2003 IEEE XIII Workshop on Neural Networks for Signal Processing (IEEE Cat. No.03TH8718)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129677868","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}
{"title":"Kernel PCA for quantization of analog vectors on a pyramid","authors":"J. Gomes, S. Mitra","doi":"10.1109/NNSP.2003.1318059","DOIUrl":"https://doi.org/10.1109/NNSP.2003.1318059","url":null,"abstract":"A kernel PCA-based method is proposed for vector quantization that performs a partition of the input space with less distortion than conventional transform coding. The distortion improvement comes at a modest increase in computational complexity and increased entropy of the quantization index stream. The proposed system is especially attractive under severe hardware constraints for which the digital hardware for entropy coding is unavailable. Numerical results are presented to validate the proposed method and to demonstrate the trade-off between distortion and entropy provided by kernel PCA at the source coding level.","PeriodicalId":315958,"journal":{"name":"2003 IEEE XIII Workshop on Neural Networks for Signal Processing (IEEE Cat. No.03TH8718)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129499863","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}
{"title":"Adaptive room acoustic response simulation: a virtual 3D application","authors":"G. Costantini, D. Casali, A. Uncini","doi":"10.1109/NNSP.2003.1318066","DOIUrl":"https://doi.org/10.1109/NNSP.2003.1318066","url":null,"abstract":"In this paper we propose a method to simulate a 3D acoustical environment in which sound sources are positioned in well defined sides. Our method is real-time applications oriented, due to the low computational cost of the implemented operations. The spatial position that the human brain assigns to a sound is influenced mainly by the differences between the sound signals that reach the listener's ears, related to the sound source angulation with respect to the listener's head. The reverberation effect, on the other side, depends on the type of environment. All this elements have to be simulated in order to give the illusion that a sound comes from a particular position in a particular environment. To obtain this result, we perform a suitable sound processing, that can be separated in two main tasks: reverberation and spatialization. The first one is mainly related to the environment itself: it depends on the shape of the environment and on the absorption coefficients of the walls. This is the most computational intensive component, if we want to reproduce it accurately, so we approximate it by an adaptive IIR filter. By the spatialization, the listener hears the sound as coming from a particular direction. This task, carried out by using the head related transfer functions (HRTFs), has to be applied to every sound source differently.","PeriodicalId":315958,"journal":{"name":"2003 IEEE XIII Workshop on Neural Networks for Signal Processing (IEEE Cat. No.03TH8718)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127930134","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}
A. Barros, J. Príncipe, Y. Takeuchi, C. H. Sales, N. Ohnishi
{"title":"An algorithm based on the even moments of the error","authors":"A. Barros, J. Príncipe, Y. Takeuchi, C. H. Sales, N. Ohnishi","doi":"10.1109/NNSP.2003.1318087","DOIUrl":"https://doi.org/10.1109/NNSP.2003.1318087","url":null,"abstract":"We propose an algorithm based on a linear combination of the even moments of the error for adaptive filtering, called weighted even moment (WEM) algorithm. It is similar to the well-known least mean square (LMS) and to the family of algorithms proposed by Walach and Widrow (1994). This later ones were shown to behave poorer than the LMS, however, when the noise was Gaussian. We study the WEM algorithm convergence behavior and deduce equations for the misadjustment and the learning time. The results showed that the WEM had better performance than the LMS when the noise had a Gaussian distribution.","PeriodicalId":315958,"journal":{"name":"2003 IEEE XIII Workshop on Neural Networks for Signal Processing (IEEE Cat. No.03TH8718)","volume":"21 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122124845","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}