Sharif Digital Repository

We propose a time-varying dual accelerated gradient method for minimizing the average of n strongly convex and smooth functions over a time-varying network with n nodes. We prove that the time-varying dual accelerated gradient ascent method converges at an R-linear rate with the time to reach an ϵ-neighborhood of the solution being of O([Formula presented]ln⁡[Formula presented]), where c is a constant depending on the graph and objective function parameters and M is a constant depending on the initial values. We test the proposed method on two classes of problems: L2-regularized least squares and logistic classification problems. For each class, we generate 1000 problems and use the... 

Transferring large models in federated learning (FL) networks is often hindered by clients' limited bandwidth. We propose $ extsf {FedAA}$ , an FL algorithm which achieves fast convergence by exploiting the regularized Anderson acceleration (AA) on the global level. First, we demonstrate that FL can benefit from acceleration methods in numerical analysis. Second, $ extsf {FedAA}$ improves the convergence rate for quadratic losses and improves the empirical performance for smooth and strongly convex objectives, compared to FedAvg, an FL algorithm using gradient descent (GD) local updates. Experimental results demonstrate that employing AA can significantly improve the performance of FedAvg,... 

A coupled statistical and conditional generative adversarial neural network is used for 3D reconstruction of both homogeneous and heterogeneous porous media from a single two-dimensional image. A statistical approach feeds the deep network with conditional data, and then the reconstruction is trained on a deep generative network. The conditional nature of the generative model helps in network stability and convergence which has been optimized through a gradient-descent-based optimization method. Moreover, this coupled approach allows the reconstruction of heterogeneous samples, a critical and serious challenge in conventional reconstruction methods. The main contribution of this work is to... 

Motion capture is a process that movements of living organisms like human or objects are captured and the results are processed for the desired applications. These applications are in rehabilitation, sports, film industry and etc. There are many techniques and instruments for motion capture that optical camera systems are the most accurate ones. But these cameras are high cost and limited to labs. Some sensors like Inertial Measurement Units (IMU) and recently, Kinect cameras have been considered by many researchers because these are low cost and easy to use. But problems like bias, accumulated error and occlusion make them look for improvements. Fusion algorithms are one of the best methods... 

Motion capture is a process that movements of living organisms like human or objects are captured and the results are processed for the desired applications. These applications are in rehabilitation, sports, film industry and etc. There are many techniques and instruments for motion capture that optical camera systems are the most accurate ones. But these cameras are high cost and limited to labs. Some sensors like Inertial Measurement Units (IMU) and recently, Kinect cameras have been considered by many researchers because these are low cost and easy to use. But problems like bias, accumulated error and occlusion make them look for improvements. Fusion algorithms are one of the best methods... 

One of the most important challenges in target parameters estimation via model-free methods is selection of the most effective input parameters namely features selection (FS). Indeed, irrelevant features can degrade the estimation performance. In the current study, the challenge of choosing among the several plant parameters is tackled by means of the innovative FS algorithm named ranking of features with minimum deviation from the target parameter (RFMD). The selected features accompanied with the stable and the fast learning algorithm of multilayer perceptron (MLP) neural network (i.e. Levenberg-Marquardt algorithm) which is a combination of gradient descent and Gauss-newton learning... 

In this paper cross-correlation of measurable/unmeasurable parameters of nuclear power plants (NPPs) are detected. Correlation techniques including Pearson's, Spearman's, and Kendall-tau give appropriate input parameters for training/prediction of the target unmeasurable parameters. Fuel and clad maximum temperatures of uncontrolled withdrawal of control rods (UWCR) transient of Bushehr nuclear power plant (BNPP) are used as the case study target parameters. Different model-free methods including decision tree (DT), feed-forward back propagation neural network (FFBPNN) accompany with different learning algorithms (i.e. gradient descent with momentum (GDM), scaled conjugate gradient (SCG),... 

In the last years, an appreciable effort has been directed toward developing empirical models to link the relative permeability of gas condensate reservoirs to the interfacial tension and velocity as well as saturation. However, these models suffer from non-universality and uncertainties in setting the tuning parameters. In order to alleviate the aforesaid infirmities in this study, comprehensive modeling was carried out by employing numerous smart computer-aided algorithms including Support Vector Regression (SVR), Least Square Support Vector Machine (LSSVM), Extreme Learning Machine (ELM), Multilayer Perceptron (MLP), Group Method of Data Handling (GMDH), and Gene Expression Programming... 

In this study, coupled Generative Adversarial and Auto-Encoder neural networks have been used to reconstruct realizations of three-dimensional porous media. The gradient-descent-based optimization method is used for training and stabilizing the neural networks. The multi-scale reconstruction has been conducted for both sandstone and carbonate samples from an Iranian oilfield. The sandstone contains inter and intra-grain porosity. The generative adversarial network predicts the inter-grain pores and the auto-encoder provides the generative adversarial network result with intra-grain pores (micro-porosity). Different matching criteria, including porosity, permeability, auto-correlation... 

In this paper, some important operating parameters of nuclear power plants (NPPs) transients are forecasted using different supervised learning methods including feed-forward back propagation (FFBP) neural networks such as cascade feed-forward neural network (CFFNN), statistical methods such as support vector regression (SVR), and localized networks such as radial basis network (RBN). Different learning algorithms, including gradient descent (GD), gradient descent with momentum (GDM), scaled conjugate gradient (SCG), Levenberg-Marquardt (LM), and Bayesian regularization (BR) are used in CFFNN method. SVR method is used with different kernel functions including Gaussian, polynomial, and... 

Transformers are the key component in power system transmission and distribution networks. Condition-based maintenance will increase their expected life, and online monitoring is essential to ensure operation reliability. In this paper, a new approach to transformer online monitoring is provided based on partial discharge (PD) measurement. Simultaneous measurements of PD using IEC60270 and radio frequency (RF) techniques are employed to explore new features that can be used to distinguish between internal PDs and external interference, as well as among different internal PD sources. Stream clustering based on the density grid method with only a few features required is used to categorize... 

This article presents a robust adaptive fractional order proportional integral derivative controller for a class of uncertain fractional order nonlinear systems using fractional order sliding mode control. The goal is to achieve closed-loop control system robustness against the system uncertainty and external disturbance. The fractional order proportional integral derivative controller gains are adjustable and will be updated using the gradient method from a proper sliding surface. A supervisory controller is used to guarantee the stability of the closed-loop fractional order proportional integral derivative control system. Finally, fractional order Duffing–Holmes system is used to verify... 

In this paper, we propose a new derivative-free preconditioned conjugate gradient method in order for solving large-scale square and under-determined nonlinear systems of equations. The proposed method is also equipped with a relaxed nonmonotone line search technique. Under some suitable assumptions, the global convergence property is established. Numerical results on some square and under-determined test systems show the efficiency and effectiveness of the new method in practice. An application of the new method for solving nonlinear integro-differential equations is also provided. © 2016 Elsevier Ltd  

In this paper the projected gradient method is applied as an effective gradient-based topology optimization algorithm in order to direct energy propagation through the desired region of composites microstructure. Rayleigh Damping model is also used in order to take the effect of internal damping mechanisms into account and thus, to fill in the gap between the designed layouts and those in reality. The success of the proposed algorithm is illustrated through several numerical experiments by revealing a set of various designed optimal layouts besides their corresponding energy distributions  

Time-dependent origin-destination (TDOD) demand is a key input of dynamic traffic assignment (DTA) in advanced traffic management systems. Model reliability is highly dependent on the accuracy of this information. One method to achieve TDOD demand matrices is to use a primary demand matrix and traffic volume counts in some links of a network. This paper proposes a bi-level model to correct the TDOD demand matrix. The extended gradient method (EGM) - an iterative method that minimizes the discrepancy between the counted and estimated traffic volumes - is a suggested means to solve this problem. The methodology is first tested on a small synthetic network to verify its performance. Then, it is... 

Based on a modified secant equation proposed by Li and Fukushima, we derive a stepsize for the Barzilai-Borwein gradient method. Then, using the newly proposed stepsize and another effective stepsize proposed by Dai et al. in an adaptive scheme that is based on the objective function convexity, we suggest a modified two-point stepsize gradient algorithm. We also show that the limit point of the sequence generated by our algorithm is first-order critical. Finally, our numerical comparisons done on a set of unconstrained optimization test problems from the CUTEr collection are presented. At first, we compare the performance of our algorithm with two other two-point stepsize gradient algorithms... 

Taking advantage of the attractive features of Hestenes-Stiefel and Dai-Yuan conjugate gradient methods, we suggest two globally convergent hybridizations of these methods following Andrei's approach of hybridizing the conjugate gradient parameters convexly and Powell's approach of nonnegative restriction of the conjugate gradient parameters. In our methods, the hybridization parameter is obtained based on a recently proposed hybrid secant equation. Numerical results demonstrating the efficiency of the proposed methods are reported  

The Burr III distribution properly approximates many familiar distributions such as Normal, Lognormal, Gamma, Weibull, and Exponential distributions. It plays an important role in reliability engineering, statistical quality control, and risk analysis models. The Burr III distribution has four parameters known as location, scale, and two shape parameters. The estimation process of these parameters is controversial. Although the maximum likelihood estimation (MLE) is understood as a straightforward method in parameters estimation, using MLE to estimate the Burr III parameters leads to maximize a complicated function with four unknown variables, where using a conventional optimization such as... 

Response surface methodology (RSM) explores the relationships between several explanatory variables and one or more response variables. The main idea of RSM is to use a set of designed experiments to obtain an optimal response. RSM tries to simplify the original problem through some polynomial estimation over small sections of the feasible area, elaborating on optimum provision through a well known optimization technique, say Gradient Method. As the real world problems are usually very complicated, polynomial estimation may not perform well in providing a good representation of the objective function. Also, the main problem of the Gradient Method, getting trapped in local minimum (maximum),... 

Following the approach proposed by Dai and Liao, we introduce two nonlinear conjugate gradient methods for unconstrained optimization problems. One of our proposed methods is based on a modified version of the secant equation proposed by Zhang, Deng and Chen, and Zhang and Xu, and the other is based on the modified BFGS update proposed by Yuan. An interesting feature of our methods is their account of both the gradient and function values. Under proper conditions, we show that one of the proposed methods is globally convergent for general functions and that the other is globally convergent for uniformly convex functions. To enhance the performance of the line search procedure, we also...