• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.24 no.10
• /
• pp.1317-1325
• /
• 2000

A new efficient numerical method for computing three-dimensional, unsteady, incompressible flows is presented. To eliminate the restriction of CFL condition, a fully-implicit time advancement in which the Crank-Nicolson method is used for both the diffusion and convection terms, is adopted. Based on an approximate block LU decomposition method, the velocity -pressure decoupling is achieved. The additional decoupling of the intermediate velocity components in the convection term is made for the fully -implicit time advancement scheme. Since the iterative procedures for the momentum equations are not required, the velocity components decouplings bring forth the reduction of computational cost. The second-order accuracy in time of the present numerical algorithm is ascertained by computing decaying vortices. The present decoupling method is applied to minimal channel flow unit with DNS (Direct Numerical Simulation).

• Journal of applied mathematics & informatics
• /
• v.28 no.3_4
• /
• pp.687-696
• /
• 2010

In this paper, an algorithm for computing the sparse approximate inverse factor of matrix $A^{T}\;A$, where A is an $m\;{\times}\;n$ matrix with $m\;{\geq}\;n$ and rank(A) = n, is proposed. The computation of the inverse factor are done without computing the matrix $A^{T}\;A$. The computed sparse approximate inverse factor is applied as a preconditioner for solving normal equations in conjunction with the CGNR algorithm. Some numerical experiments on test matrices are presented to show the efficiency of the method. A comparison with some available methods is also included.

• Proceedings of the Korean Nuclear Society Conference
• /
• 1997.10a
• /
• pp.85-90
• /
• 1997

Previous our numerical results in computing point kinetics equations show a possibility in developing approximations to estimate sensitivity responses of nuclear reactor We recalculate sensitivity responses by maintaining the corrections with first order of sensitivity parameter. We present a method for computing sensitivity responses of nuclear reactor based on an approximation derived from point kinetics equations. Exploiting this approximation, we found that the first order approximation works to estimate variations in the time to reach peak power because of their linear dependence on a sensitivity parameter, and that there are errors in estimating the peak power in the first order approximation for larger sensitivity parameters. To confirm legitimacy of our approximation, these approximate results are compared with exact results obtained from our previous numerical study.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.25 no.3
• /
• pp.537-547
• /
• 2001

The paper proposes the fuzzy logic global approximate optimization strategies in optimal sizing of automotive A-pillar trim with rib structures for occupant head protection. Two different strategies referred to as evolutionary fuzzy modeling (EFM) and neuro-fuzzy modeling (NFM) are implemented in the context of global approximate optimization. EFM and NFM are based on soft computing paradigms utilizing fuzzy systems, neural networks and evolutionary computing techniques. Such approximation methods may have their promising characteristics in a case where the inherent nonlinearity in analysis model should be accommodated over the entire design space and the training data is not sufficiently provided. The objective of structural design is to determine the dimensions of rib in A-pillar, minimizing the equivalent head injury criterion HIC(d). The paper describes the head-form modeling and head impact simulation using LS-DYNA3D, and the approximation procedures including fuzzy rule generation, membership function selection and inference process for EFM and NFM, and subsequently presents their generalization capabilities in terms of number of fuzzy rules and training data.

• Journal of IKEEE
• /
• v.26 no.4
• /
• pp.736-741
• /
• 2022

Lattice-based cryptography is the most practical post-quantum cryptography because it enjoys strong worst-case security, relatively efficient implementation, and simplicity. Ring learning with errors (R-LWE) is a public key encryption (PKE) method of lattice-based encryption (LBC), and the most important operation of R-LWE is the modular polynomial multiplication of rings. This paper proposes a method for optimizing modular multipliers based on approximate computing (AC) technology, targeting the medium-security parameter set of the R-LWE cryptosystem. First, as a simple way to implement complex logic, LUT is used to omit some of the approximate multiplication operations, and the 2's complement method is used to calculate the number of bits whose value is 1 when converting the value of the input data to binary. We propose a total of two methods to reduce the number of required adders by minimizing them. The proposed LUT-based modular multiplier reduced both speed and area by 9% compared to the existing R-LWE modular multiplier, and the modular multiplier using the 2's complement method reduced the area by 40% and improved the speed by 2%. appear. Finally, the area of the optimized modular multiplier with both of these methods applied was reduced by up to 43% compared to the previous one, and the speed was reduced by up to 10%.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2003.05a
• /
• pp.386-393
• /
• 2003

For fast Cholesky factorization, it is most important to reduce the number of non-zero elements by ordering methods. Minimum deficiency ordering produces less non-zero elements. However, since it is very slow. the minimum degree algorithm is widely used. To improve the computation time, Rothberg's AMF uses an approximate deficiency instead of computing the deficiency. In this paper we present simple efficient methods to obtain a good approximate deficiency using information related to cliques. Experimental results show that our proposed method produces better ordering quality than that of AMF.

• Journal of Semiconductor Engineering
• /
• v.1 no.1
• /
• pp.31-37
• /
• 2020

We present an approximate DRAM architecture for energy-efficient deep learning. Our key premise is that by bounding memory errors to non-critical information, we can significantly reduce DRAM refresh energy without compromising recognition accuracy of deep neural networks. To validate the key premise, we make extensive Monte-Carlo simulations for several well-known convolutional neural networks such as LeNet, ConvNet and AlexNet with the input of MINIST, CIFAR-10, and ImageNet, respectively. We assume that the highest-order 8-bits (in single precision) and 4-bits (in half precision) are protected from retention errors under the proposed architecture and then, randomly inject bit-errors to unprotected bits with various bit-error-rates. Here, recognition accuracies of the above convolutional neural networks are successfully maintained up to the 10^-5-order bit-error-rate. We simulate DRAM energy during inference of the above convolutional neural networks, where the proposed architecture shows the possibility of considerable energy saving up to 10 ~ 37.5% of total DRAM energy.

• Journal of Internet Computing and Services
• /
• v.14 no.1
• /
• pp.63-69
• /
• 2013

Approximate Frequent pattern mining is to find approximate patterns, not exact frequent patterns with tolerable variations for more efficiency. As the size of database increases, much faster mining techniques are needed to deal with huge databases. Moreover, it is more difficult to discover exact results of mining patterns due to inherent noise or data diversity. In these cases, by mining approximate frequent patterns, more efficient mining can be performed in terms of runtime, memory usage and scalability. In this paper, we study the characteristics of an approximate mining algorithm based on probabilistic technique and run performance evaluation of the efficient approximate frequent pattern mining algorithm. Finally, we analyze the test results for more improvement.

• Journal of the Korean Society for Railway
• /
• v.16 no.5
• /
• pp.372-378
• /
• 2013

This paper presents an approximate model for computing the peak demand power in a metro railway system. The peak demand of substations can be calculated using the current vector iteration method. But the existing method requires many repeated calculations to determine the peak demand power, which makes it difficult to apply to the real-time peak power control problem. In this paper, we assume that none of the conditions vary except source impedance and make an approximate model for rapid calculation based on changes in the impedance of the power substation. The proposed model result is approximately the same as the existing model, which is demonstrated through simulation.

• IEIE Transactions on Smart Processing and Computing
• /
• v.6 no.4
• /
• pp.299-304
• /
• 2017

A wireless passive sensor network is a network consisting of sink nodes, sensor nodes, and radio frequency (RF) sources, where an RF source transfers energy to sensor nodes by radiating RF waves, and a sensor node transmits data by consuming the received energy. Against theoretical expectations, a wireless passive sensor network suffers from many practical difficulties: scarcity of energy, non-simultaneity of energy reception and data transmission, and inefficiency in allocating time resources. Perceiving such difficulties, we propose a simple contending-type medium access control (MAC) scheme for many sensor nodes to deliver packets to a sink node. Then, we derive an approximate expression for the network-wide throughput attained by the proposed MAC scheme. Also, we present an approximate expression for the optimal partition, which maximizes the saturated network-wide throughput. Numerical examples confirm that each of the approximate expressions yields a highly precise value for network-wide throughput and finds an exactly optimal partition.