Least square matching (LSM) is one of the most accurate image matching methods in photogrammetry and remote sensing. The main disadvantage of the LSM is its high computational complexity due to large size of observation equations. To address this problem, in this paper a novel method, called fast Least square matching (FLSM) is being presented. The main idea of the proposed FLSM is decreasing the size of the observation equations to improve the efficiency of the matching process. For this purpose, the pixels in the matching window are ordered using a special robustness measure. Then, a specific percent of the pixels with the highest robustness is selected for matching process. The phase congruency and entropy measures are used to compute the proposed robustness measure. The proposed FLSM method was successfully applied to match various synthetic and real image pairs, and the results demonstrate its capability to increase matching efficiency. The matching results show that the proposed FLSM method is three times faster than standard LSM method.

While the Hammerstein expression is computationally attractive for modeling of nonlinear systems, the optimal calculation of filter coefficients is practically cumbersome. A proper solution to this problem is the use of adaptive algorithms. In this paper, the Hammerstein Normalized LMS algorithm is proposed by deriving the corresponding time varying optimal step-size parameter in a closed form. The convergence behavior of this algorithm is inspected using computer simulations. The results show that the proposed algorithm achieves a faster convergence speed compared to the Hammerstein LMS algorithm, practically at the cost of an acceptable increase in computations.

In seepage problems, the coefficients of permeability in Laplace equation are usually assumed to be constant vs. both space and time; but in reality these coefficients are variable. In this study, the effect of material deformation due to external loads (consolidation) and variation of head in the consolidation process are considered. For the first case, formulation of kx and ky can be defined by a second order binominal equation in order to take into account the material changes due to volume changes. For the second case, kx and ky can be defined as a function of unknown total head. The solution of the resulting non-linear differential equation is found using the Least square Finite Element formulation. In order to increase the accuracy of the solution, eight nodal (isoperimetric) elements were obtained. This method was used satisfactorily to solve several seepage problems and to examine the accuracy and convergence of the results. The effect of a variable coefficient of permeability may not be significant on small dams, but as the height of the dam increases, the effect becomes more considerable. It is believed that a variable permeability analysis such as the one described in this paper should be taken into account.

The conventional fuzzy Least-squares time series models show undesirable performance when the fuzzy data set involves the outliers. By introducing a strategy to detect the outliers, this paper introduced a method for reducing the influence of outliers on the future predictions. For this purpose, according to the weighted square distance error, an estimation procedure was suggested for determining the exact coefficients in the presence of outliers. The parameters of the fuzzy time series model were then estimated using an iterative algorithm. In order to identify the potential outliers of the fuzzy data, a  quality control chart was employed based on the center of gravity criterion of fuzzy data. The defuzzification method was also employed to examine the performance of the proposed method via some  scatter plots. Several common goodness-of-fit criteria used in traditional time series models were also extended to compare the performance of the proposed fuzzy time series method to an existing method. The effectiveness of the proposed method was illustrated through two numerical examples including a simulation study. The results clearly indicated that the proposed model performs well in terms of the both scatter plot criteria and goodness-of-fit evaluations in cases where the potential outliers exist among the fuzzy data.

A meshless approach, collocation discrete Least square (CDLS) method, is extended in this paper, for solving elasticity problems. In the present CDLS method, the problem domain is discretized by distributed field nodes. The field nodes are used to construct the trial functions. The moving Least-squares interpolant is employed to construct the trial functions. Some collocation points that are independent of the field nodes are used to form the total residuals of the problem. The Least-squares technique is used to obtain the solution of the problem by minimizing the summation of the residuals for the collocation points. The final stiffness matrix is symmetric and therefore can be solved via efficient solvers. The boundary conditions are easily enforced by the penalty method. The present method does not require any mesh so it is a truly meshless method. Numerical examples are studied in detail, which show that the present method is stable and possesses good accuracy, high convergence rate and high efficiency.

Intelligent control of real control problems based on reinforcement learning often requires decision-making in a large or continuous state-action space. Since the number of adjustable parameters in discrete reinforcement learning has a direct relationship with cardinality of the state-action space of the problem, so in such problems, we are faced with the curse of dimensiality, low learning speed and low efficiency. The use of continuous reinforcement learning methods to overcome these problems have attracted many research interests. In this paper a novel Neural Reinforcement Learning (NRL) scheme is proposed. The presented method is model free and learning rate independent, and is obtained by combining Least squares Policy Iteration (LSPI) with Radial Basis Functions (RBF) as a function approximator, and we call it "Neural Least squares Policy Iteration" (NLSPI). In this method, by using the basis functions defined in the RBF neural network structure, we have provided a solution to solve the challenge of defining the state-action basis functions in LSPI. In order to validate the presented method, the performance of the proposed algorithm in solving two control problems has been compared with other methods. The overall results show the superiority of our method in learning the pseudo-optimal policy.

A dual-time implicit mesh-less method is presented for unsteady compressible flow calculations. Polynomial Least-square (PLS) and Taylor series Least-square (TLS) procedures are used to estimate the spatial derivatives at each node and their computational efficiencies are compared. Also, the effect of the neighbor stencil selection on the accuracy of the method is investigated. As a new approach, different neighboring stencils are used for the highly stretched point distribution inside the boundary layer region and the inviscid isotropic point distribution outside this area. The unsteady flows over stationary and moving objects at subsonic and transonic flow conditions are solved. Results indicate the computational efficiency of the method in comparison with the alternative approaches. The convergence histories of the flow solution show that the PLS method is computationally faster than TLS method. In addition, the eight point neighboring stencil inside the viscous region is more efficient than other choices.

Many meshless methods are presented in recent years for solving partical differential equations. All of them use a background mesh to introducing the Guossian points for solution of the integral equation. Numerical integration has significant effects on convergence and accuracy of solution from many meshless methods. They are meshless only from the point of view of interpolation of the field variables. In this paper we present a fully meshless procedure for linear and nonlinear convection-diffusion equation problems in steady and transient forms. In this new method a fully Least squares method is used in both function approximation and the discretization of the governing differential equations. The discritized equations are obtained via a discrete Least squares method in which the sum of the squared residuals are minimized with respect to unknown nodal parameters in the inner and boundary nodes. In this process no numerical integration is needed and the obtained equations are symmetric and positive definite. The shape functions are derived using the moving Least squares (MLS) method of function approximation with a spline weighting function. The convergency and accuracy of the method is programmed using visual Fortran 6.5and accuracy of the results are compared and verified with reference to some examples.