Stability and encoding properties of two-layer nonlinear feedback neural networks are examined. Bidirectionality is introduced in neural nets to produce two-way associative search for stored associations. The bidirect...
详细信息
Stability and encoding properties of two-layer nonlinear feedback neural networks are examined. Bidirectionality is introduced in neural nets to produce two-way associative search for stored associations. The bidirectional associative memory (BAM) is the minimal two-layer nonlinear feedback network. The author proves that every n-by-p matrix M is a bidirectionally stable heteroassociative content-addressable memory for both binary/bipolar and continuous neurons. When the BAM neutrons are activated, the network quickly evolves to a stable state of two-pattern reverberation, or resonance. The stable reverberation corresponds to a system energy local minimum. Heteroassociative information is encoded in a BAM by summing correlation matrices. The BAM storage capacity for reliable recall is roughly m>min (n,p). It is also shown that it is better on average to use bipolar (-1,1) coding than binary
During the past few years there has been an increasing interest in applying higher -order statistics, namely cumulants, and their associated Fourier transforms, polyspectra, to a wide range of signalprocessing and sy...
详细信息
This paper proposes an unidirectional ring systolic architecture for implementing the hidden Markov models (HMMs). This array architecture maximizes the strength of VLSI in terms of intensive and pipelined computing a...
详细信息
The algebraic properties of binary image algebra (BIA) are used to mathematically describe and compare symbolic substitution and pattern recognition algorithms for implementation on optical processors.
The algebraic properties of binary image algebra (BIA) are used to mathematically describe and compare symbolic substitution and pattern recognition algorithms for implementation on optical processors.
A learning algorithm based on temporal difference of membrane potential of the neuron is proposed for self-organizing neural networks. It is independent of the neuron nonlinearity, so it can be applied to analog or bi...
详细信息
A noncausal autoregressive (NCAR) plus additive noise model is presented for model-based spectrum estimation of two-dimensional sinusoidal signals in noise. The maximum-likelihood (ML) procedure provides consistent an...
详细信息
A noncausal autoregressive (NCAR) plus additive noise model is presented for model-based spectrum estimation of two-dimensional sinusoidal signals in noise. The maximum-likelihood (ML) procedure provides consistent and efficient parameter estimates for NCAR models with bilateral neighbor sets, and these properties carry over to the maximum-likelihood estimates of parameters for Gaussian-NCAR-plus-noise models. By assuming a toroidal lattice the complexity of the ML equation is significantly reduced with little impact on the observed accuracy of the estimated spectra. Initial conditions for starting the ML computation are proposed. Experimental results are presented for various signal-to-noise ratios.< >
The incoherent optical neuron (ION) subtracts inhibitory inputs from excitatory inputs optically by utilizing separate device responses. Those factors that affect the operation of the ION are discussed, including nonl...
详细信息
The incoherent optical neuron (ION) subtracts inhibitory inputs from excitatory inputs optically by utilizing separate device responses. Those factors that affect the operation of the ION are discussed, including nonlinearity of the inhibitory element, input noise, device noise, system noise and crosstalk. A computer simulation of these effects is performed on a version of S.A. Ellias and S. Grossberg's (1975) on-center off-surround competitive neural network.< >
A time-series consisting of sinusoids observed in additive i.i.d. noise or in additive colored Gaussian noise of unknown spectral density is considered. The number of harmonics, as well as their amplitudes and frequen...
详细信息
A time-series consisting of sinusoids observed in additive i.i.d. noise or in additive colored Gaussian noise of unknown spectral density is considered. The number of harmonics, as well as their amplitudes and frequencies are determined using the one-dimensional diagonal slice of the fourth-order cumulant. Applications to the detection of cubic phase coupling are discussed.< >
A lattice version of the recursive instrumental variable method for adaptive parameter identification of ARMA (autoregressive moving-average) processes is developed. Appropriate choice of the instrumental variables le...
详细信息
A lattice version of the recursive instrumental variable method for adaptive parameter identification of ARMA (autoregressive moving-average) processes is developed. Appropriate choice of the instrumental variables leads to cumulant-based AR parameter estimates. Cumulant-based normal equations may be obtained by using nonconventional orthogonality conditions in the linear prediction problem. The development leads to a pair of lattices, one excited by the observed process y(n), and the other by the instrumental process z(n). The lattices are coupled through order-update and time-update equations. The lattice structure yields the AR compensated residual time series. Hence, adaptive versions of cumulant-based MA parameter identification algorithms are directly applicable. Some convergence results are presented.< >
One-dimensional cumulants are useful for solving a wide variety of system-theoretic and signalprocessing problems. How can 1D cumulants be calculated? This paper shows how to compute 1D third- and fourth- order cumul...
详细信息
One-dimensional cumulants are useful for solving a wide variety of system-theoretic and signalprocessing problems. How can 1D cumulants be calculated? This paper shows how to compute 1D third- and fourth- order cumulants for a single-input, single-output linear time-invariant system that is described by a state-variable model. Results are given for both stationary and nonstationary systems. Emphasis is placed on computational procedures that are recursive. For stationary systems, our algorithms are recursive with respect to the cumulant lag. Because the cumulant is nonsymmetric with respect to lag zero, different algorithms are needed for positive and negative lags. For nonstationary systems, our algorithms are recursive either in time or lag, or in both time and lag. Our recursive algorithms are presented in two ways, namely matix form or Kronecker product form. The Kronecker product form is especially useful for fourth-order cumulant calculations.
暂无评论