• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.23 no.11
• /
• pp.1357-1363
• /
• 2019

In this paper, we compares the performance of the gredient descent optimizers of the Faster Region-based Convolutional Neural Network (R-CNN) model for the chromosome object detection in digital images composed of human metaphase chromosomes. In faster R-CNN, the gradient descent optimizer is used to minimize the objective function of the region proposal network (RPN) module and the classification score and bounding box regression blocks. The gradient descent optimizer. Through performance comparisons among these four gradient descent optimizers in our experiments, we found that the Adamax optimizer could achieve the mean average precision (mAP) of about 52% when considering faster R-CNN with a base network, VGG16. In case of faster R-CNN with a base network, ResNet50, the Adadelta optimizer could achieve the mAP of about 58%.

• Physical Therapy Korea
• /
• v.29 no.3
• /
• pp.208-214
• /
• 2022

Background: Pronated foot posture (PFP) contributes to excessive dynamic knee valgus (DKV). Although foot orthoses such as medial arch support (MAS) are widely and easily used in clinical practice and sports, few studies have investigated the effect of MAS on the improvement of DKV during stair descent in individuals with a PFP. Moreover, no studies reported the degree of improvement in DKV according to the severity of PFP when MAS was applied. Objects: This study aimed to examine the immediate effect of MAS on DKV during stair descent and determine the correlation between navicular drop distance and changes in DKV when MAS is applied. Methods: Twenty individuals with a PFP (15 males and five females) participated in this study. The navicular drop test was used to measure PFP severity. The frontal plane projection angle (FPPA) was calculated under two conditions, with and without MAS application, using 2-dimensional video analysis. Results: During stair descent, the FPPA with MAS (173.1° ± 4.7°) was significantly greater than that without MAS (164.8° ± 5.8°) (p < 0.05). There was also a significant correlation between the navicular drop distance and improvement in the FPPA when MAS was applied (r = 0.453, p = 0.045). Conclusion: MAS application can affect the decrease in DKV during stair descent. In addition, MAS application should be considered to improve the knee alignment for individuals with greater navicular drop distance.

• International Journal of Control, Automation, and Systems
• /
• v.3 no.2
• /
• pp.244-251
• /
• 2005

In this paper, we study the problem of stabilizing a general nonlinear control system by means of gradient descent control method which is a dynamic feedback control law. In this method, the general nonlinear control system can be considered as an affine nonlinear control systems. Then by using Sontag's formula we investigate the stability (asymptotic) of the general nonlinear control system.

• Korean Journal of Air-Conditioning and Refrigeration Engineering
• /
• v.15 no.6
• /
• pp.495-501
• /
• 2003

Control algorithms for the absorption air conditioning system may be developed by using dynamic models of the system. The simplified effective dynamic models, which can predict the dynamic behaviors of the system, may help to develop effective control algorithms for the system. In this study, control algorithms for an absorption air conditioning system were developed by using a dynamic simulation program. A cooling water inlet temperature control algorithm, a chilled water outlet temperature control algorithm, and a supply air temperature control algorithm, were developed and analyzed. The steepest descent method was used as an optimal algorithm. The simulation results showed energy savings and the effective controls of an absorption air conditioning system.

• Proceedings of the Korean Society of Propulsion Engineers Conference
• /
• 2011.11a
• /
• pp.254-257
• /
• 2011

Two types of thrusters(Descent Control Thruster (DCT) for reducing landing speed and Attitude Control Thruster (ACT) for attitude control) are mounted on the propulsion system of Ground test model lunar lander. In this paper, plume impingement effect and ground effect between DCT Modules are analyzed using numerical method when the impact occurred close to the ground.

• Transactions on Control, Automation and Systems Engineering
• /
• v.4 no.4
• /
• pp.330-340
• /
• 2002

In this Paper. a non-linear approach to a design of model reference adaptive control is presented. The approach is demonstrated by a case study of a simple single-pole and no zero, linear, discrete-time plant. The essence of the idea is to generate a full non-linear model of the plant dynamics and the parameter adaptation dynamics as a gradient descent algorithm with respect to a Riemannian metric. It is shown how a Riemannian metric can be chosen so that the modelled plant dynamics do in fact match the true plant dynamics. The performance of the proposed scheme is compared to a traditional model reference adaptive control scheme using the classical sensitivity derivatives (Euclidean gradients) for the descent algorithm.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.58 no.1
• /
• pp.23-27
• /
• 2009

The joint dispatch of regional electricity markets can improve the overall economic efficiency of interconnected markets by increasing the combined social welfare of the interconnected markets. This paper presents a new decentralized optimization technique based on Augmented Lagrangian Relaxation (ALR) to perform the joint dispatch of interconnected electricity markets. The Block Coordinate Descent (BCD) technique is applied to decompose the inseparable quadratic term of the augmented Lagrangian equation into individual market optimization problems. The Interior Point/Cutting Plane (IP/CP) method is used to update the Lagrangian multiplier in the decomposed market optimization problem. The numerical example is presented to validate the effectiveness of the proposed decentralized method.

• Journal of applied mathematics & informatics
• /
• v.24 no.1_2
• /
• pp.179-190
• /
• 2007

In this paper, we consider the general variational inequality GVI(F, g, C), where F and g are mappings from a Hilbert space into itself and C is the fixed point set of a nonexpansive mapping. We suggest and analyze a new modified hybrid steepest-descent method of type method $u_{n+l}=(1-{\alpha}+{\theta}_{n+1})Tu_n+{\alpha}u_n-{\theta}_{n+1g}(Tu_n)-{\lambda}_{n+1}{\mu}F(Tu_n),\;n{\geq}0$. for solving the general variational inequalities. The sequence $\{x_n}\$ is shown to converge in norm to the solutions of the general variational inequality GVI(F, g, C) under some mild conditions. Application to constrained generalized pseudo-inverse is included. Results proved in the paper can be viewed as an refinement and improvement of previously known results.

• 제어로봇시스템학회:학술대회논문집
• /
• 2000.10a
• /
• pp.210-210
• /
• 2000

In this paper, we have investigated the properties of multi-layer perceptron (MLP) for odour patterns classification and concentration estimation simultaneously. When the MLP may be has a fast convergence speed with small error and excellent mapping ability for classification, it can be possible to use for classification and concentration prediction of volatile chemicals simultaneously. However, the conventional MLP, which is back-Propagation of error based on the steepest descent method, was difficult to use for odour classification and concentration estimation simultaneously, because it is slow to converge and may fall into the local minimum. We adapted the Levenberg-Marquardt(LM) algorithm [4,5] having advantages both the steepest descent method and Gauss-Newton method instead of the conventional steepest descent method for the simultaneous classification and concentration estimation of odours. And, We designed the artificial odour sensing system(Electronic Nose) and applied LM-BP algorithm for classification and concentration prediction of VOC gases.

• The Journal of the Korea Contents Association
• /
• v.4 no.3
• /
• pp.99-105
• /
• 2004

This paper proposes optimal learning rates in the gradient descent training of multilayer perceptrons, which are a separate learning rate for weights associated with each neuron and a separate one for assigning virtual hidden targets associated with each training pattern Effectiveness of the proposed error function was demonstrated for a handwritten digit recognition and an isolated-word recognition tasks and very fast learning convergence was obtained.