The Infona portal uses cookies, i.e. strings of text saved by a browser on the user's device. The portal can access those files and use them to remember the user's data, such as their chosen settings (screen view, interface language, etc.), or their login data. By using the Infona portal the user accepts automatic saving and using this information for portal operation purposes. More information on the subject can be found in the Privacy Policy and Terms of Service. By closing this window the user confirms that they have read the information on cookie usage, and they accept the privacy policy and the way cookies are used by the portal. You can change the cookie settings in your browser.
On large datasets, the popular training approach has been stochastic gradient descent (SGD). This paper proposes a modification of SGD, called averaged SGD with feedback (ASF), that significantly improves the performance (robustness, accuracy, and training speed) over the traditional SGD. The proposal is based on three simple ideas: averaging the weight vectors across SGD iterations, feeding the averaged...
Networked systems are common and crucial. One of the canonical problems in such systems is distributed resource allocation. From this rather broad class of problems, we consider a convex non-smooth resource allocation problem with a global resource constraint. Specifically, the objective function is separable and consists of a sum of convex functions, each associated with a node in a given network...
The traditional searching scheme of independent component analysis (ICA) is based on gradient algorithm. And a learning step size is required beforehand. It couldn't resolve the problem of convergence. To overcome the drawback, an improved particle swarm optimization (PSO) is applied to ICA algorithm. Firstly, the dynamic inertia weight which is based on evolution speed and aggregation degree is introduced...
In this paper, we present a new algorithm of multi-point iterative method for simultaneous determination of all roots of polynomial. Its convergence was researched. The computation is carried out by simple steepest descent rule with adaptive variable step-size. The specific examples illustrated that the proposed method can find simultaneously the roots of polynomials at a very rapid convergence and...
The present paper considers distributed consensus algorithms for agents evolving on a connected compact homogeneous (CCH) manifold. The agents track no external reference and communicate their relative state according to an interconnection graph. The paper first formalizes the consensus problem for synchronization (i.e. maximizing the consensus) and balancing (i.e. minimizing the consensus); it thereby...
We consider the design of optimal static feedback gains for interconnected systems subject to architectural constraints on the distributed controller. These constraints are in the form of sparsity requirements for the feedback matrix, which means that each controller has access to information from only a limited number of subsystems. We derive necessary conditions for the optimality of structured...
In this paper we present a gradient method to iteratively update local controllers of a distributed linear system driven by stochastic disturbances. The control objective is to minimize the sum of the variances of states and inputs in all nodes. We show that the gradients of this objective can be estimated distributively using data from a forward simulation of the system model and a backward simulation...
An observer-based Hamiltonian identification algorithm for quantum systems has been recently proposed by two of the authors to estimate the dipole moment matrix of a quantum system requiring the measurement of the populations on all states. This could be experimentally difficult to achieve. We propose here an extension to a 3-level quantum system, having access to the population of the ground state...
In practice, the convergence rate and stability of perturbation based extremum-seeking (ES) schemes can be very sensitive to the curvature of the plant map. This sensitivity arises from the use of a gradient descent adaptation algorithm. Such ES schemes may need to be conservatively tuned in order to maintain stability over a wide range of operating conditions, resulting in slower optimisation than...
Maximum likelihood estimation(MLE) is widely applied in system identification because it is consistent and has excellent convergence properties. However gradient based optimization of likelihood function might end up in local convergence. It is known that for ARMAX and ARARX models, providing a large enough Signal-to-Noise-Ratio(SNR) will avoid the potential local convergence. We show the same condition...
Fast convergence-rate, low computation complexity and good stability are important goals in the researching area of neural network learning algorithm. A kind of parallel computing lagged-start hybrid optimization algorithm is studied, it not only integrates the basic gradient method and the unconstrained optimization algorithm to realize the supplement of their advantages, but also makes full use...
A method for optimization of continuous nonlinear functions is introduced. Seed Throwing Optimization is a probabilistic metaheuristic. It has roots in hill climbing and the evolutionary computation like technique harmony search. The relationship to these algorithms is shown in this paper. Our method is tested in a benchmark and compared to other metaheuristics. Seed Throwing Optimization is a randomized...
A novel adaptive nonlinear controller is presented for nonlinear active noise control systems, which is expanded by memory function mapping on the basis of a single neuron structure, and a generalized filtered-X gradient descent algorithm is developed to attenuate the nonlinear, non-Gaussian noises, which defines the weighted sum of Renyi's quadratic error entropy and the mean square error as the...
Snakes or active contours are widely used in the fields of computer vision and image processing. Gradient vector flow (GVF) is an effective external force for active contours, but it is based on isotropic diffusion and doesn't take the image structure into account. In this study, a novel external force for active contours, named edge preserving gradient vector flow is proposed for active contours...
In this study, we propose a new chaotic global optimization method using the Lagrangian method to solve a nonlinear constrained optimization problem. Firstly, we explain the convergence behavior of the first order method regarding convexity of the Lagrangian with respect to decision variables in terms of linear stability theory. Further, we propose a new optimization method in which the convergence...
In order to solve the difficult problem that how to reduce the overshot and shorten the regulating time of the PID controller based on BP neural network, a self-tuning PID controller based on improved BP neural network is presented. The parameters of the PID controller are calculated by an improved BP Neural Network according to the input and output and the error of the PID controller. It is introduced...
This paper investigates the reason why the predistorted signal is sampled and feeds back to an adaptation block as reference for digital predistorter (DPD) to be updated in real-time (RT) by comparing a non-real-time (NRT) method with off-time processing. In addition, by deriving a simple algorithm based on gradient descent algorithm with minimum square error criterion, the steady-state condition...
In this paper, we study distributed classification of targets in a large scale sensor network setting. Specifically, we consider sensor nodes which can measure only a part of the feature vector and whose communication capabilities are limited to only their neighbouring nodes. We formulate a distributed classification algorithm that learns the optimal (large-margin) hyperplane separating the two classes,...
In order to improve the accuracy, robustness, and computational load of c-means clustering models, a series of hybrid solutions have been proposed. Mixtures of fuzzy (FCM) and possibilistic c-means (PCM) clustering generally attempted to avoid the noise sensitivity of the former and the coincident clusters of the latter. On the other hand, mixtures of fuzzy and hard c-means (HCM) have been proposed...
In this paper, we investigate a class of unconstrained minimization methods including Fletcher-Reeves (abbr. FR) conjugate gradient method with perturbations. Their stepsizes are determined by generalized Wolfe line search. We prove that these methods are globally convergent under mild conditions, and in doing so, we remove various boundedness conditions such as boundedness from blow of f, boundedness...
Set the date range to filter the displayed results. You can set a starting date, ending date or both. You can enter the dates manually or choose them from the calendar.