Abstract-An orthogonal forward selection (OFS) algorithm based on leave-one-out (LOO) misclassification rate is proposed for the construction of radial basis function (RBF) classifiers with tunable units. Each stage of the construction process determines a RBF unit, namely its centre vector and diagonal covariance matrix as well as weight, by minimising the LOO statistics. This OFS-LOO algorithm is computationally efficient and it is capable of constructing parsimonious RBF classifiers that generalise well. Moreover, the proposed algorithm is fully automatic and the user does not need to specify a termination criterion for the construction process. The effectiveness of the proposed RBF classifier construction procedure is demonstrated using three classification benchmark examples.
Abstract-Blind and semiblind adaptive schemes are proposed for joint maximum likelihood (ML) channel estimation and data detection for multiple-input multiple-output (MIMO) systems. The joint ML optimisation over channel and data is decomposed into an iterative two-level optimisation loop. An efficient global optimisation search algorithm called the repeated weighted boosting search is employed at the upper level to identify the unknown MIMO channel model while an enhanced ML sphere detector called the optimised hierarchy reduced search algorithm aided ML detector is used at the lower level to perform the ML detection of the transmitted data. A simulation example is included to demonstrate the effectiveness of these two schemes.
Abstract-Semi-blind space-time equalisation is considered for dispersive multiple-input multiple-output systems that employ high-throughput quadrature amplitude modulation signalling. A minimum number of training symbols, approximately equal to the dimension of the space-time equaliser (STE), are first utilised to provide a rough initial least squares estimate of the STE's weight vector. A gradient-Newton-type concurrent constant modulus algorithm and soft decision-directed scheme is then applied to adapt the STE. The proposed semi-blind adaptive STE is capable of converging fast and accurately to the optimal minimum mean square error STE solution.
I. INTRODUCTIONWith the aid of smart antenna arrays and by exploiting both the space and time dimensions, space-time processing is capable of effectively improving the achievable system capacity, coverage and quality of service by suppressing both intersymbol interference and co-channel interference [1]- [7]. In this contribution, we consider space-division multiple-access (SDMA) induced frequency selective multiple-input multipleoutput (MIMO) systems that employ quadrature amplitude modulation (QAM) signalling. A bank of space-time equalisers (STEs) [8]-[14] form the multiuser receiver. Adaptive implementation of STE can be realised using the training-based least mean square (LMS) or recursive least squares (RLS) algorithm [15]. However, a large number of training symbols is required to adapt a STE, which considerably reduces the achievable system throughput. Blind adaptive methods may be applied to adjust a STE, which does not require training symbols and, therefore, does not reduce the achievable system throughput. However, blind methods require high computational complexity and, moreover, they result in unavoidable estimation and decision ambiguities [16], [17], which can only be resolved with the aid of a few training symbols. At the cost of requiring a few training symbols, semi-blind schemes can avoid the estimation and decision ambiguity problem and are computationally simpler than their blind counterparts.Many semi-blind methods have been proposed for narrowband MIMO systems [18]-[24]. In particular, the work of [24] has developed a semi-blind spatial equalisation scheme for narrowband MIMO systems that employ QAM signalling. In this semi-blind method, a few training symbols, approximately equal to the dimension of the spatial equaliser, are first used to provide a rough least squares (LS) estimate of the spatial equaliser's weight vector. The stochastic-gradient (SG) based constant modulus algorithm (CMA) and soft decision-directed (SDD) scheme, originally developed for blind equalisation of
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.