Scientia Iranica
Volume:24 Issue: 3, 2017

Learning invariant features while the distribution of training (source domain) and test (target domain) sets are different is crucial. However, most of the dimensionality reduction methods perform poorly in facing with domain shift problem either in original or latent spaces. In this paper we introduce DiReT, a Discriminative Dimensionality Reduction approach for multi-source Transfer learning, that aims at constructing a latent space across domains in a semi-supervised manner. Our main contribution is to reduce the drift in distributions across domains and concurrently preserve the separability between classes. DiReT by establishing a bridge between source and target domains guarantees the knowledge transformation along different domains. Empirical evidences indicate that DiReT manages to improve substantially over dimensionality reduction methods, especially with extracting more features from multiple domains. We evaluate DiReT against other well-known dimensionality reduction and transfer learning methods on three synthetic and two real world datasets.

The paper proposes an adaptive Frequency Domain based Switching Median Filter (FDSMF) for the restoration of images corrupted by periodic noise. The proposed algorithm incorporates region-growing technique to effectively identify noisy peak areas of the Fourier transformed image in to a binary noise map image. The restoration phase of the algorithm replaces the corrupted frequencies with the median of uncorrupted frequencies by recursive median filter. Experimental results from different naturally and artificially corrupted images at various noise levels / types reveal that the performance of the proposed algorithm in restoring images corrupted by periodic noise is better than other competing algorithms in terms of subjective and objective metrics.

For a set of colored points, a region is called extit{color-spanning} if it contains at least one point of each color. In this paper, we first consider the problem of maintaining extit{the smallest color-spanning interval} for a set of $n$ points with $k$ colors on the real line such that the insertion and deletion of an arbitrary point takes $O(log^2 n)$ worst-case time. Then, we exploit the data structure to show that there is an $O(nlog^2 n)$ time algorithm to compute extit{the smallest color-spanning square} for a set of $n$ points with $k$ colors in the plane. This is a new way to improve the $O(nk log n)$ time algorithm presented by Abellanasetal~citeAbellanas01smallestcolor-spanning} when $k=omega(log n)$. We also consider the problem of computing the smallest color-spanning square in a special case in which we have at most two points from each color. We present an $O(nlog n)$ time algorithm to solve the problem which improves the result presented by Arkinetal~citeArkin} by a factor of $log n$

Finding two disjoint simple paths on two given sets of points is a geometric problem introduced by Jeff Erickson. This problem has various applications in computational geometry, like robot motion planning, generating polygon etc. We will present a reduction from planar Hamiltonian path to this problem, and prove that it is NP-Complete. To the best of our knowledge, no study has considered its complexity up until now. We also present a reduction from planar Hamiltonian path problem to the problem of “finding a path on given points in the presence of arbitrary obstacles” and prove that it is NP-Complete too. Also, we present a heuristic algorithm with time complexity of O(n^4) to solve this problem. The proposed algorithm first calculates the convex hull for each of the entry points and then produces two simple paths on the two entry point sets.

In this paper, application of the surface current method (SCM) for analyzing and optimization of electromechanical devices is proposed. SCM is one of the numerical techniques in electromagnetic field analysis. In SCM only magnetic boundaries are subdivided against Finite Element Method (FEM) which is subdivided all of domain therefore the calculation resource of SCM is much lesser than FEM. SCM with low calculation resource is one of the best numerical techniques for magnetic devices optimization. In this paper, using SCM three electromechanical systems have been optimized based on minimization of weight per force. Verification of simulation results is done by FEM.

OFDM is an e ective multicarrier transmission technique with one primary disadvantage; it su ers from high Peak-to-Average Power Ratio (PAPR). Although clipping and ltering is a simple and e ective method for PAPR reduction, it makes in-band and out-of-band noise, which degrades the bit error rate performance and spectral eciency. Publications on this subject show that clipped samples could be reconstructed at the receiver by using oversampled signal and bandwidth expansion. By building on published literature, this paper aims to achieve a low-complexity method. The proposed method has complexity order of O(L2) to solve linear system, where L indicates the number of clipped samples. Simulation results con rm that our proposed method leads to both a better biterror rate performance and a lower complexity than similar methods. These results also show that our method o ers adequate performance, especially at low clipping ratios.

Vedic mathematics, is an ancient computing methodology, has a unique computational technique based on 16 sutras (formulae). These formulae can be directly applied for the optimization of the algebraic computation. A new algorithm for the computation of the decimals of the inverse based on such ancient mathematics is reported in this paper. Sahayaks (auxiliary fraction) sutra has been used for the hardware implementation of the decimals of the inverse. On account of the Vedic formulae, reciprocal approximation of a numbers can generate “on the fly” either the n first exact decimals of the inverse, this n being either arbitrary large, or, at least, in almost all cases, 6. The reported algorithm has been implemented and functionality has been checked in T-Spice. Performance parameters like propagation delay, dynamic switching power consumption are calculated through spice-spectre of 90nm CMOS technology. The propagation delay of the resulting 4-digit reciprocal approximation algorithm was only ~1.8uS and consumes ~24.7mW power. The implementation methodology offered substantial reduction of propagation delay, and dynamic switching power consumption from its counterpart (NR) based implementation.

Contingency ranking is one of the most important stages in the analysis of power system security. In this paper, an integrated algorithm has been proposed to address this issue. This algorithm employs neural networks method to fast-estimate the power system parameters and stochastic frontier analysis (SFA) in order to calculate the efficiency of each contingency. Network security indices (voltage violation and line flow violation indices) and economic indices (locational marginal price and congestion cost indices) have been simultaneously considered to rank the contingencies. The efficiency of each contingency shows its severity and indicates that it affects network security and economic indices concurrently. The proposed algorithm has been tested on IEEE 14-bus and 30-bus test power systems. Simulation results show the high efficiency of the algorithm.Test results indicate that predicted quantities are estimated accurately and quickly. The proposed method is capable of producing fast and accurate network security indices and economic indices, so it can be used for online ranking.

Trac sign recognition can be performed in two phases i.e. (1) detection and (2) recognition; detection deals with sensing a trac sign in real world image or video frame while recognition is aboutreading its contents. A trac signs database may contain samples with varying font sizes and stylesused for printing the interior of a trac sign and the contents may also be shifted away from the center of gravity. In this paper, we utilize the energy compaction property of Discrete Cosine Transform (DCT) to propose a trac sign recognition (TSR) system which can generate invariant features for varying font styles, scaled up, scaled down and translated contents of a sign. Experiments on synthetic and real world images datasets show that the features generated by our proposed method have great intraclass similarity and interclass variation. We have also shown that our proposed method outperforms Eigen based recognition method [1] and is comparable with the histogram of oriented gradient (HOG) approach [2] using support vector machine (SVM) classi er.

Hysteresis motors are used in special applications such as gyroscope, centrifuge, and machine tool spindle drives due to their unique features such as synchronism, self-starting and developing smooth torque. Dynamic modeling of hysteresis motors is essential to predict of the transient performance, studying the dynamic stability and development of modern closed-loop control and estimation strategies. This paper develops a new dynamic modeling for a high-speed, circumferential-flux type hysteresis motor in synchronous dq reference frame. Major previous models use fixed parameters for rotor’s equivalent circuit parameters corresponding to the major B-H loop of rotor material, whereas operational B-H loop is significantly affected by stator voltage and load torque. In this paper, a time varying dynamic model is developed, in which the parameters of the equivalent circuit of hysteresis rotor material are adjusted based on operational B-H loop. For this purpose and based on elliptical assumption for B-H loops, the hysteresis lag angle β is updated due to the applied stator voltage and available load torque. Developed mathematical model satisfies many theoretical aspects of hysteresis motor behavior in transient and steady-state situations. The model offers a tool for studying the start-up of hysteresis motor, change of stator voltage, variation of load torque, frequency tracking for variable-speed applications and transient-state response to design parameters. Some simulations are provided to demonstrate the validity of developed model in Matlab/ Simulink environment and are verified via some experimental results. Proposed results verify the advantages of this model rather than previous works.

This paper investigates a novel nonsingular fast terminal sliding-mode control method for the stabilization of the uncertain time-varying and nonlinear thirdorder systems. The designed disturbance observer satis es the nite-time convergence of the disturbance approximation error and the suggested nite-time stabilizer assures the presence of the switching behavior around the switching curve in the nite time. Furthermore, this approach can overcome the singularity problem of the fast terminal sliding-mode control technique. Moreover, knowledge about the upper bounds of the disturbances is not required and the chattering problem is eliminated. Usefulness and e ectiveness of the o ered procedure are con rmed by numerical simulation results.

In this paper, it is shown that repeating average filter increases the uniform patterns of noisy textures and consequently increases the classification accuracy of textures. In other words, for noisy textures,first, an average filter such as 3x3 mean filter is applied to each image then a feature extraction method such as LBP is used to extract features of filtered image. The more value of noise the more repeating ofaverage filter should be applied to textures. It is true that repeating average filter decreases the variance of noisy image. However, in this paper it is shownthat by repeating average filter for textures the variance of texture decreases then increases. So,average filter must be repeated while the variance of image decreases and until the variance is increased, it must be stop. Using convolution to apply average filter for an image takes so much time, therefore a simple technique is proposed in this paper that increases the speed of average filtering significantly.After noise reduction, by using LBP operator,features of texture areextracted for classification. Implementationson Outex, CUReT and UIUC datasets determine that the performance of proposed method is higher than some advanced noise resistant LBP variants such as BRINT and CRLBP.

This work considers the possibility of applying electrochemical noise analysis to fuel cell diagnostics. Theoretical hypothesis and experimental result have shown that spectral characteristics of electrical uctuations depend on water management processes inside the PEMFC. It has been established that the spectrum of electrical uctuations in low frequency range has the nature of flicker noise. The frequency ranges of 0.1-1 Hz convenient for single cell as well as for stack diagnostics are revealed. The results show that the proposed approach can be considered as an eff ective tool to diagnose of fuel cells, namely allowing for prediction of drying and flooding.

In this paper, we propose a threshold increasing algorithm for a (t;n) lattice based threshold multi-stage secret sharing (TMSSS) scheme. For realization of the changeability feature, we use the zero addition protocol to construct a new (tœ;n) TMSSS scheme. Therefore, the new scheme enjoys the significant feature of threshold changeability in addition to the inherited features of being multi-stage, multi-use and verifiable from our previously proposed lattice based TMSSS scheme. Furthermore, we use the improved TMSSS scheme to propose a threshold decryption algorithm for the learning with error (LWE) based public key encryption scheme based on Lindner and Peikert’s. For threshold decryption, each authorized subset of participants decrypts the ciphertext partially and sends the result to the combiner. Using them, the combiner can decrypt the ciphertext. The security of both schemes is based on the hardness of lattice problems, i.e., LWE and inhomogeneous small integer solution (ISIS) problems, which are believed to resist against the quantum algorithms. The proposed schemes are efficient, especially in the participants’ side, making them suitable for the applications in which the participants have limited processing capacities.

This paper presents a new control strategy for an Optimal Unified Power Flow Controller (OUPFC) through a Lyapunov energy function in terms of local parameters to improve the transient stability of a power system. The OUPFC is a hybrid configuration of Flexible AC Transmission System (FACTS) devices, i.e., an arrangement of small-size Unified Power Flow Controller (UPFC) and a full-scale Phase Shifting Transformer (PST). In this study, a new term of OUPFC’s energy function and its injection model in a simplified structure preserving model is developed and implemented in a two-machine power system using MATLAB/Simpower. The ability of the OUPFC controller to enhance the transient stability is compared to that of UPFC. The results show that using the proposed control strategy for OUPFC, leads to more first swing oscillations mitigation and stability margin enlargement. Also it is concluded that by using OUPFC with appropriate angles in proper locations, compensation of UPFC''s angle displacement may come true. So comparing OUPFC to UPFC, it can be said that, OUPFC enjoys another degree of freedom.

This paper presents a new robust fault detection and isolation scheme using fuzzy wavelet network based on the bounded error approach. An efficient hybrid design algorithm which consists of the orthogonal least square and the artificial bee colony algorithm is proposed to design fuzzy wavelet network for modeling normal and faulty behavior of the system. The proposed model provides an alternative description of the behavior of the system with high accuracy, but it suffers from model uncertainty, because of model-reality mismatch in practical applications. To overcome this difficulty, the bounded error approach inspired from robust identification theory, is applied to estimate the model uncertainty which defines a confidence interval of the model output and derives adaptive threshold for residual evaluation. Also, online fault isolation process is performed using fuzzy wavelet network models of the faulty system and analyzing the relation between a bank of residuals. Performance and efficiency of the proposed scheme is evaluated by simulating the nonlinear two-tank liquid level control system. Finally, some performance indexes are defined and then the Monte-Carlo analysis is carried out to evaluate the reliability and robustness of the proposed scheme.