• Journal of the Korean Data and Information Science Society
• /
• v.28 no.6
• /
• pp.1245-1255
• /
• 2017

The online learning is a process of obtaining the solution for a given objective function where the data is accumulated in real time or in batch units. The stochastic gradient descent method is one of the most widely used for the online learning. This method is not only easy to implement, but also has good properties of the solution under the assumption that the generating model of data is homogeneous. However, the stochastic gradient method could severely mislead the online-learning when the homogeneity is actually violated. We assume that there are two heterogeneous generating models in the observation, and propose the a new stochastic gradient method that mitigate the problem of the heterogeneous models. We introduce a robust mini-batch optimization method using statistical tests and investigate the convergence radius of the solution in the proposed method. Moreover, the theoretical results are confirmed by the numerical simulations.

• Journal of Electrical Engineering and Technology
• /
• v.9 no.1
• /
• pp.80-89
• /
• 2014

A stochastic optimal power flow (S-OPF) model considering uncertainties of load and wind power is developed based on chance constrained programming (CCP). The difficulties in solving the model are the nonlinearity and probabilistic constraints. In this paper, a limit relaxation approach and an iterative learning control (ILC) method are implemented to solve the S-OPF model indirectly. The limit relaxation approach narrows the solution space by introducing regulatory factors, according to the relationship between the constraint equations and the optimization variables. The regulatory factors are designed by ILC method to ensure the optimality of final solution under a predefined confidence level. The optimization algorithm for S-OPF is completed based on the combination of limit relaxation and ILC and tested on the IEEE 14-bus system.

• Journal of Institute of Control, Robotics and Systems
• /
• v.6 no.3
• /
• pp.263-272
• /
• 2000

In this paper we give some geometric condition for a stochastic nonlinear system and we propose a control method for a stochastic nonlinear system using neural networks. Since a competitive learning neural networks has been developed based on the stochastcic approximation method it is regarded as a stochastic recursive filter algorithm. In addition we provide a filtering and control condition for a stochastic nonlinear system called the perfect filtering condition in a viewpoint of stochastic geometry. The stochastic nonlinear system satisfying the perfect filtering condition is decoupled with a deterministic part and purely semi martingale part. Hence the above system can be controlled by conventional control laws and various intelligent control laws. Computer simulation shows that the stochastic nonlinear system satisfying the perfect filtering condition is controllable and the proposed neural controller is more efficient than the conventional LQG controller and the canonical LQ-Neural controller.

• Journal of Platform Technology
• /
• v.9 no.1
• /
• pp.3-14
• /
• 2021

Cognitive Radio network (CR) is a promising paradigm that helps the unlicensed user (Secondary User) to analyse the spectrum and coordinate the spectrum access to support the creation of common control channel (CCC). The cooperation of secondary users and broadcasting between them is done through transmitting messages in CCC. In case, if the control channels may get jammed and it may directly degrade the network's performance and under such scenario jammers will devastate the control channels. Hopping sequences may be one of the predominant approaches and it may be used to fight against this problem to confront jammer. The jamming attack can be alleviated using one of the game modelling approach and in this proposed scheme stochastic games has been analysed with more single users to provide the flexible control channels against intrusive attacks by mentioning the states of each player, strategies ,actions and players reward. The proposed work uses a modern player action and better strategic view on game theoretic modelling is stochastic game theory has been taken in to consideration and applied to prevent the jamming attack in CR network. The selection of decision is based on Q learning approach to mitigate the jamming nodes using the optimal MDP decision process

• Journal of the Korean Institute of Telematics and Electronics C
• /
• v.35C no.3
• /
• pp.79-89
• /
• 1998

The Error Back-Propagation(EBP) algorithm is widely applied to train a multi-layer perceptron, which is a neural network model frequently used to solve complex problems such as pattern recognition, adaptive control, and global optimization. However, the EBP is basically a gradient descent method, which may get stuck in a local minimum, leading to failure in finding the globally optimal solution. Moreover, a multi-layer perceptron suffers from locking a systematic determination of the network structure appropriate for a given problem. It is usually the case to determine the number of hidden nodes by trial and error. In this paper, we propose a new algorithm to efficiently train a multi-layer perceptron. OUr algorithm uses stochastic perturbation in the weight space to effectively escape from local minima in multi-layer perceptron learning. Stochastic perturbation probabilistically re-initializes weights associated with hidden nodes to escape a local minimum if the probabilistically re-initializes weights associated with hidden nodes to escape a local minimum if the EGP learning gets stuck to it. Addition of new hidden nodes also can be viewed asa special case of stochastic perturbation. Using stochastic perturbation we can solve the local minima problem and the network structure design in a unified way. The results of our experiments with several benchmark test problems including theparity problem, the two-spirals problem, andthe credit-screening data show that our algorithm is very efficient.

• ETRI Journal
• /
• v.28 no.1
• /
• pp.59-66
• /
• 2006

Learning behaviors of a radial basis function network (RBFN) using a singular value decomposition (SVD) and stochastic gradient (SG) algorithm, together named RBF-SVD-SG, for odor sensing systems are analyzed, and a fast training method is proposed. RBF input data is from a conducting polymer sensor array. It is revealed in this paper that the SG algorithm for the fine-tuning of centers and widths still shows ill-behaving learning results when a sufficiently small convergence coefficient is not used. Since the tuning of centers in RBFN plays a dominant role in the performance of RBFN odor sensing systems, our analysis is focused on the center-gradient variance of the RBFN-SVD-SG algorithm. We found analytically that the steadystate weight fluctuation and large values of a convergence coefficient can lead to an increase in variance of the center-gradient estimate. Based on this analysis, we propose to use the least mean square algorithm instead of SVD in adjusting the weight for stable steady-state weight behavior. Experimental results of the proposed algorithm have shown faster learning speed and better classification performance.

• Journal of the Korean Institute of Telematics and Electronics C
• /
• v.34C no.7
• /
• pp.82-91
• /
• 1997

In this paper, we analyze the dynamics of langevine competitive learning neural network based on its fokker-plank equation. From the viewpont of the stochastic differential equation (SDE), langevine competitive learning equation is one of langevine stochastic differential equation and has the diffusin equation on the topological space (.ohm., F, P) with probability measure. We derive the fokker-plank equation from the proposed algorithm and prove by introducing a infinitestimal operator for markov semigroups, that the weight vector in the particular simplex can converge to the globally optimal point under the condition of some convex or pseudo-convex performance measure function. Experimental resutls for pattern recognition of the remote sensing data indicate the superiority of langevine competitive learning neural network in comparison to the conventional competitive learning neural network.

• International journal of advanced smart convergence
• /
• v.12 no.2
• /
• pp.56-66
• /
• 2023

The goal of deep learning is to extract complex features from multidimensional data use the features to create models that connect input and output. Deep learning is a process of learning nonlinear features and functions from complex data, and the user data that is employed to train deep learning models has become the focus of privacy concerns. Companies that collect user's sensitive personal information, such as users' images and voices, own this data for indefinite period of times. Users cannot delete their personal information, and they cannot limit the purposes for which the data is used. The study has designed a deep learning method that employs privacy protection technology that uses distributed collaborative learning so that multiple participants can use neural network models collaboratively without sharing the input datasets. To prevent direct leaks of personal information, participants are not shown the training datasets during the model training process, unlike traditional deep learning so that the personal information in the data can be protected. The study used a method that can selectively share subsets via an optimization algorithm that is based on modified distributed stochastic gradient descent, and the result showed that it was possible to learn with improved learning accuracy while protecting personal information.

• 제어로봇시스템학회:학술대회논문집
• /
• 1992.10b
• /
• pp.312-317
• /
• 1992

This paper proposes a new framework of an autonomous and distributed flexible manufacturing system - Multi Client Robot Groups(MCR) - and describes a stochastic learning scheme applied to managerial problems of the system. The MCR is composed of groups of manufacturing robots, named Client Robots (CRs), which are capable of both versatility and independence in their performances. The MCR is expected to have high performance because the MCR can perform concurrent and corporative processing. However, the system performance is determined by the organizations of the CR groups. Therefore the treatment of the managerial problems and organizations of the system are important problems. In this paper, it is assumed that CR groups being able to processing tasks are selected stochastically based on the strengths of the robot groups. The learning scheme adjusting the strength is introduced to organize the groups in the system and control the each performance of the groups according to the total system performance. Finally, some experimental results of the learning scheme are shown.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.31B no.7
• /
• pp.211-218
• /
• 1994

Stochastic computation employs random pulse streams to represent numbers. In this paper, we study a new method to implement the number system which uses the ratio of the numbers of ones and zeros in the pulse streams. In this number system. if P is the probability that a pulse is one in a pulse stream then the number X represented by the pulse stream is defined as P/(1-P). We propose circuits to implement the basic operations such as addition multiplication and sigmoid function with this number system and examine the error characteristics of such operations in stochastic computation. We also propose a neuron model and derive a learning algorithm based on backpropagation for the 3-layered feedforward neural networks. We apply this learning algorithm to a digit recognition problem. To analyze the results, we discuss the errors due to the variance of the random pulse streams and the quantization noise of finite length register.