Convergence Analysis of Zero Attracting Natural Gradient Non-Parametric Maximum Likelihood Algorithm

Performance analysis of the deficient length augmented CLMS algorithm for second order noncircular complex signals

In digital communication system that the Modified Constant Modulus Algorithm (MCMA) reduced the use of the adaptive equalization algorithm to combat the Inter-symbol Interference (ISI). MCMA is relatively brief operation. The major point of MCMA that it only achieves moderate convergence rate and steady state mean square error (MSE). In this paper suggest, MCMA equalization improve the performance with parallel structure. It combines Modified Constant Modulus Algorithm(MCMA) and Modified Decision Directed(MDD) algorithm. By exploiting the inherent structural relationship between the 4-QAM signal's coordinates and 16-QAM signal's coordinates, another style of cost function for Modified Constant Modulus Algorithm(MCMA) is defined and If it happen to offset of received signals and MCMA is poor performance in order to overcome this because the paper combines apply for MCMA and MDD(Modified Decision Direct) algorithm. By computer simulation, we confirmed that the proposed PMCMA-MDD algorithm has the fater convergence rate and steady mean square error than the conventional MCMA.

A least mean square (LMS) algorithm with clipped data is studied for use when updating the weights of an adaptive filter with correlated Gaussian input. Both stationary and nonstationary environments are considered. Three main contributions are presented. The first, corresponding to the stationary case, is a proof of the convergence of the algorithm in the case of a M-dependent sequence of correlated observation vectors. It is proven that the steady state mean square misalignment of the adaptive filter weights has an upper bound proportional to the algorithm step size mu . The second contribution, also belonging to the stationary case, is the derivation of the expressions of convergence time N/sub c/ and steady state mean square excess estimation error epsilon . It is shown that N/sub c/ is proportional to 1/( mu lambda ), with lambda being the minimum eigenvalue of the input covariance matrix. It is also shown that the product N/sub c/ epsilon is independent of mu . For a given epsilon , the convergence time increases with the eigenvalue spread of the input covariance matrix and the filter length, as well as its input noise power. The range of mu that achieves tolerable values of N/sub c/ and epsilon is determined. The third contribution is concerned with the nonstationary case. It is shown that the mean square excess estimation error is the sum of the two terms with opposite dependencies on mu . An optimum value of mu is derived.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>

The paper is concerned with analyzing the roundoff error effect on the tracking performance of the block least mean square (BLMS) algorithm when used in the adaptive identification of a time-varying plant. Rounding quantization is assumed. Expressions are derived for the steady state mean square error, steady state mean square weight deviation, and the corresponding optimum step sizes. It is found that the mean square error and the mean square weight deviation are decreasing functions of both the filter coefficients wordlength and the algorithm block size. Expressions for minimum and maximum block lengths are derived. The performance of the BLMS is compared to that of the conventional LMS algorithm. It is found that the BLMS possesses a higher resistance to roundoff errors than the LMS algorithm. The theoretical results of the paper are validated by computer simulations.

Decision Feedback Equalizer (DFE) offers a good compromise between complexity and performance for channels with strong Intersymbol Interference (ISI). However, their blind or Decision-Directed (DD) adaptation may endure the inherent limitation from blind equalization algorithm or DD algorithm, i.e., 1) the blind equalization such as Constant Modulus Algorithm (CMA) has the ¿eye-opened¿ ability, but moderate convergence rate and steady state Mean Square Error (MSE); 2) the DD algorithm has faster convergence rate and lower steady state MSE, but it always cannot work unless the MSE of received signals is very low. So for overcoming above disadvantages we study the geometrical relation between CMA1-2 and DD algorithm and obtain the method of soft switch between CMA1-2 and DD algorithm. Consequently a new DFE based on CMA1-2 and DD algorithm is proposed. Based on theoretical analysis and computer simulations, the proposed approach is shown to be much more effective than the conventional DFE.

Deep in-memory architecture (DIMA) provides considerable latency and throughput improvements over conventional digital architectures. DIMA reads multiple bits per bit-line (BL) in each cycle, and performs mixed-signal processing at the periphery of the bit cell array (BCA). Past work [1] used adaptive filtering as a means to analyse the impact of nonlinear read operation on gradient descent methods implemented on DIMA. Gradient descent techniques are typically implemented in block fashion in DIMA since write operations have high overhead. This work uses block least mean square (BLMS) adaptive filters as an example to study the impact of nonlinearity and variations encountered in DIMA. The insights gained from BLMS adaptive filters are used to realize recursive least square (RLS) adaptive filters. The steady state mean square error (MSE) of both filters remains unaffected while convergence rate and stability issues can arise when these filters are implemented on DIMA. The insights are useful in context of learning systems employing gradient descent techniques.

Distributed sparse diffusion estimation based on set membership and affine projection algorithm

This paper presents two classes of adaptive blind algorithms based on second- and higher order statistics. The first class contains fast recursive algorithms whose cost functions involve second and third- or fourth-order cumulants. These algorithms are stochastic gradient-based but have structures similar to the fast transversal filters (FTF) algorithms. The second class is composed of two stages: the first stage uses a gradient adaptive lattice (GAL) while the second stage employs a higher order-cumulant (HOC) based least mean squares (LMS) filter. The computational loads for these algorithms are all linearly proportional to the number of taps used. Furthermore, the second class, as various numerical examples indicate, yields very fast convergence rates and low steady state mean square errors (MSE) and intersymbol interference (ISI). MSE convergence analyses for the proposed algorithms are also provided and compared with simulation results. >

A convex combination LMS (least mean square) algorithm based on Krylov subspace transform is proposed in this paper. In this approach, impulse response of the unknown system is firstly transformed into Krylov subspace, in which the system structure is changed into sparse. Then an improved proportionate normalized LMS (IPNLMS) algorithm and a variable tap-length normalized LMS (VTNLMS) algorithm are combined. Simulations are performed to show the convergence performance of the combined algorithm. The results show that both a fast convergence rate and small steady state mean square deviation (MSD) are obtained.

A generalized performance evaluation method is presented for extended Kalman filter based state estimation of first order nonlinear dynamic systems. The first order nonlinear dynamic systems are placed in three categorizes in terms of the upper bound of the absolute values of the derivatives of the nonlinear functions. According to the proposed theoretical analysis, it is shown that different state estimation performances are achieved by applying the extended Kalman filter to the first order nonlinear systems in these three categorizes. The steady state mean square error performance of the filter is presented explicitly in terms of the bounds on the nonlinearity and noise variance. The advantage of this performance evaluation method is that the convergence property of estimation error variance can be easily concluded from the system nonlinearity before its implementation. This provides insight into what is going to happen in applications, e.g. for dynamic systems used in chaotic synchronization. The simulation of several different nonlinear dynamic systems shows the effectiveness of the proposed performance evaluation criteria.

In this paper, we investigate the convergence performance of a sparsified kernel least mean square (KLMS) algorithm in which the input is added into the dictionary only when the prediction error in amplitude is larger than a preset threshold. Under certain conditions, we derive an approximate value of the steady-state excess mean square error (EMSE). Simulation results confirm the theoretical predictions and provide some interesting findings, showing that the sparsification can not only be used to constrain the network size (hence reduce the computational burden) but also be used to improve the steady-state performance in some cases.

In partial diffusion-based least mean square (PDLMS) scheme, each node shares a part of its intermediate estimate vector with its neighbors at each iteration. In this paper, besides studying the general PDLMS scheme, we figure out how the noisy links deteriorate the network performance during the exchange of weight estimates. We investigate the steady state mean square deviation (MSD) and derive a theoretical expression for it. We also derive the mean and mean-square convergence conditions for the PDLMS algorithm in the presence of noisy links. Our analysis reveals that unlike the PDLMS with ideal links, the steady-state network MSD performance of the PDLMS algorithm is not improved as the number of entries communicated at each iteration increases. Strictly speaking, the noisy links condition imposes more complexity to the MSD derivation that has a noticeable effect on the overall performance. This term violates the trade-off between the communication cost and the estimation performance of the networks in comparison with the ideal links. Our simulation results substantiate the effect of noisy links on PDLMS algorithm and verify the theoretical findings. They match well with theory.

The adaptive kernel algorithms usually achieve a good convergence performance and a tracking performance due to the universal approximator, offering an excellent solution to many problems with nonlinearities. However, as is well known, the convergence rate and steady-state error of adaptive filtering algorithm are a pair of inherent contradictions, and the kernel method is not exceptional. For this problem, a robust kernel adaptive filtering algorithm, called the variable-scaling factor kernel fractional lower power adaptive filtering algorithm based on the Sigmoid function, is developed by creating a new framework of cost function which combines the kernel fractional low power error criterion with the Sigmoid function for system identification of different noise environments. This new cost framework incorporates a scaling factor into the cost function of the Sigmoid kernel fractional lower power adaptive filtering algorithm (VS-SKFLP) in this paper. One of the main features in the new framework is its scaling factor. This scaling factor is used to control the steepness of the Sigmoid function, and the steepness can affect the convergence speed of filtering algorithm. The scaling factor provides a tradeoff between the convergence rate and the steady-state mean square error (MSE), which improves the convergence rate under the same steady-state mean square error. However, it is also an important problem to choose an appropriate scale factor. Therefore, a variable-scale factor SKFLP algorithm is also proposed to improve the convergence rate and steady-state MSE, simultaneously. The proposed variable-scale factor structure consists of a function of error, featuring the adaptive updates of their parameter estimated by making discerning use of the error. In this paper, the nonlinear saturation characteristic of the Sigmoid function and low order norm criterion are used to overcome the performance degradation of training data destroyed by non-Gaussian impulse noise and colored noise. Through the convergence analysis, the parameter estimation sequence of our proposed algorithm proves convergent. Simulation results show that the proposed algorithm (VS-SKFLP) outperforms other kernel adaptive filtering algorithms in system recognition with different noise environments.

This paper presents a novel variable step-size normalized PBS_LMS algorithm for adaptive filters. The fixed step-size PBS_LMS algorithm, which significantly decreases the number of calculations for updating tap-weight vector and increases the speed of convergence rate in comparison with conventional LMS algorithm, has proposed previously. However, the fixed step-size PBS_LMS algorithm as fixed step-size LMS algorithm usually results in a trade-off between the residual error and the convergence speed of the algorithm. Now in this paper the properties of Normalized LMS algorithm are used in the conventional PBS_LMS algorithm to approach the Normalized PBS_LMS algorithm with fast convergence rate. Then variable step-size is used parallel with the Normalized PBS_LMS algorithm to minimize the steady state mean square error. The function of mean square error variation is used to detecting the rate of convergence for increasing the step-size parameter to approach this goal. The computer simulations validate that the Normalized PBS_LMS algorithm can approach the faster convergence rate than the PBS_LMS algorithm. In addition, these simulations show the lower mean square error and tracking ability in Variable Step-Size Normalized PBS_LMS algorithm in comparison with the Normalized PBS_LMS algorithm.

A robust adaptive filtering algorithm based on the convex combination of two adaptive filters under the maximum correntropy criterion (MCC) is proposed. Compared with conventional minimum mean square error (MSE) criterion-based adaptive filtering algorithm, the MCC-based algorithm shows a better robustness against impulsive interference. However, its major drawback is the conflicting requirements between convergence speed and steady-state mean square error. In this letter, we use the convex combination method to overcome the tradeoff problem. Instead of minimizing the squared error to update the mixing parameter in conventional convex combination scheme, the method of maximizing the correntropy is introduced to make the proposed algorithm more robust against impulsive interference. Additionally, we report a novel weight transfer method to further improve the tracking performance. The good performance in terms of convergence rate and steady-state mean square error is demonstrated in plant identification scenarios that include impulsive interference and abrupt changes.