• 전자공학회논문지B
• /
• 제33B권2호
• /
• pp.159-167
• /
• 1996

In this paper, a new competitive learning algorithm called dynamic competitive learning (DCL) is presented. DCL is a supervised learning mehtod that dynamically generates output neuraons and nitializes weight vectors from training patterns. It introduces a new parameter called LOG (limit of garde) to decide whether or not an output neuron is created. In other words, if there exist some neurons in the province of LOG that classify the input vector correctly, then DCL adjusts the weight vector for the neuraon which has the minimum grade. Otherwise, it produces a new output neuron using the given input vector. It is largely learning is not limited only to the winner and the output neurons are dynamically generated int he trining process. In addition, the proposed algorithm has a small number of parameters. Which are easy to be determined and applied to the real problems. Experimental results for patterns recognition of remote sensing data and handwritten numeral data indicate the superiority of dCL in comparison to the conventional competitive learning methods.

• 한국공작기계학회논문집
• /
• 제14권1호
• /
• pp.37-43
• /
• 2005

Plant pipe has a fatigue life which is induced by repeated stress come from the variation of temperature and pressure. To avoid the fatigue crack of plant pipe which is produced by long term repeated stress, plant operator has to limit the mont of operating transient. This paper introduced the study result about discrimination methodology of plant transient by using the fuzzy pattern recognition. As result of applying the fuzzy pattern recognition to actual plant operation data, it is confirmed that fuzzy pattern recognition methodology can be useful for the comparison of similarity for the transients of similar output but has different time pattern.

• 한국지능시스템학회논문지
• /
• 제17권7호
• /
• pp.970-976
• /
• 2007

본 연구에서는 퍼지 추론 메커니즘에 기반 한 다항식 네트워크 패턴 분류기(Polynomial Network Pattern Classifier; PNC)를 설계하고 Particle Swarm Optimization 알고리즘을 이용하여 PNC 파라미터, 즉, 학습률, 모멘텀 계수, FCM 클러스터링의 퍼지화 계수(fuzzification Coefficient)를 최적화한다. 제안된 PNC 구조는 FCM 클러스터링에 기반한 분할 함수를 활성 함수로 사용하며, 다항식 함수로 구성된 연결가중치를 사용함으로서 기존 신경회로망 분류기의 선형적인 특성을 개선한다. PNC 구조는 언어적 해석관점에서 "If-then"의 퍼지 규칙으로 표현되며 퍼지 추론 메커니즘에 의해 구동된다. 즉 조건부, 결론부, 추론부 세 가지의 기능적 모듈로 나뉘어 네트워크 구조가 형성된다. 조건부는 FCM 클러스터링을 사용하여 입력 공간을 분할하고, 결론부는 분할된 로컬 영역을 다항식 함수로 표현한다. 마지막으로, 네트워크의 최종출력은 추론부의 퍼지추론에 의한다. 제안된 PNC는 다항식 기반 구조의 퍼지 추론 특성으로 인해 출력 공간상에 비선형 판별 함수(nonlinear discernment function)가 생성되어 분류기로서의 성능을 높인다.

• 대한전자공학회논문지
• /
• 제18권4호
• /
• pp.34-42
• /
• 1981

PMOS집적기술을 이용하여 512-Bit mask programmable ROM을 설계하고 제작하였다. ROM의 내용은 제작공정에서 gate pattern으로 기억시켰으며 chip의 출력을 512(32×16)개의 점의 행렬로써 오실로스코프에 나타내어 확인하였다. 제작된 chip은 -6V와 - l2V의 범위에서 정상적으로 동작하였다 소모전력과 전달지연시간은 -6V에서 각각 3mW와 13μsec였다. -12V에서는 소모전력이 27mW로 증가하였으며 전달지연시간은 3μsec로 감소하였다. Chip의 출력은 TTL gate의 인력을 직접 구동시킬 수 있었으며 chip select에 의하여 출력을 disable 시켰을 때는 높은 임피던스 상태를 유지하였다.

• 대한전기학회논문지
• /
• 제40권2호
• /
• pp.224-233
• /
• 1991

The multilayer expanson of single layer NN (Neural Network) was needed to solve the linear seperability problem as shown by the classic example using the XOR function. The EBP (Error Back Propagation ) learning rule is often used in multilayer Neural Networks, but it is not without its faults: 1)D.Rimmelhart expanded the Delta Rule but there is a problem in obtaining Ca from the linear combination of the Weight matrix N between the hidden layer and the output layer and H, wich is the result of another linear combination between the input pattern and the Weight matrix M between the input layer and the hidden layer. 2) Even if using the difference between Ca and Da to adjust the values of the Weight matrix N between the hidden layer and the output layer may be valid is correct, but using the same value to adjust the Weight matrixd M between the input layer and the hidden layer is wrong. Recognitron III was proposed to solve these faults. According to simulation results, since Recognitron III does not learn the three layer NN itself, but divides it into several single layer NNs and learns these with learning patterns, the learning time is 32.5 to 72.2 time faster than EBP NN one. The number of patterns learned in a EBP NN with n input and output cells and n+1 hidden cells are 2**n, but n in Recognitron III of the same size. [5] In the case of pattern generalization, however, EBP NN is less than Recognitron III.

• 한국정보전자통신기술학회논문지
• /
• 제16권6호
• /
• pp.465-471
• /
• 2023

본 논문은 CMOS Image Sensor(CIS)에 사용되는 single-slope ADC(SS-ADC)의 노이즈와 출력의 지연을 개선한 비교기 구조를 제안한다. 노이즈와 출력의 지연 특성을 개선하기 위해 비교기의 첫 번째 단의 출력 노드와 두 번째 단의 출력 노드 사이에 커패시터를 삽입하여 miller effect를 이용한 비교기 구조를 설계하였다. 제안하는 비교기 구조는 작은 capacitor를 이용하여 노이즈와 출력의 지연 및 layout 면적을 개선하였다. Single slop ADC에서 사용되는 CDS 카운터는 T-filp flop과 bitwise inversion 회로를 사용하여 설계하였고 전력 소모와 속도가 개선되었다. 또한 single slop ADC는 analog correlated double sampling(CDS)와 digital CDS를 함께 동작하는 dual CDS를 수행한다. Dual CDS를 수행함으로써 fixed pattern noise(FPN), reset noise, ADC error를 줄여 이미지 품질이 향상된다. 제안하는 comparator 구조가 사용된 single-slope ADC는 0.18㎛ CMOS 공정으로 설계되었다.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1991년도 추계학술대회 논문집 학회본부
• /
• pp.39-42
• /
• 1991

The Application of Artificial Neural Network(ANN) to forecast a load in a power system is investigated. The load forecasting is important in the electric utility industry. This technique, methodology based on the fact that parallel structure can process very fast much information is a promising approach to a load forecasting. ANN that is highly interconnected processing element in a hierachy activated by the each input. The load pattern can be divided distinctively into two patterns, that is, weekday and weekend. ANN is composed of a input layer, several hidden layers, and a output layer and the past data is used to activate input layer. The output of ANN is the load forecast for a given day. The result of this simulation can be used as a reference to a electric utility operation.

• 한국통신학회논문지
• /
• 제22권2호
• /
• pp.392-400
• /
• 1997

In this paper we porposed WCHF-fSDF filter to obtain a roration and scale-invariant correlation output. WCHF-fSDF filter is synthesized by each single CHF exttracted from scale-changed and wavelet tranformed imagesfor a refereence image as tranining images. The wavelet transform is defined as the correlation of an input image with a wavelet function. Therefore two 4f optical correlation systems are needed for pattern recognition using wavelet transform. We here include the wavelet function for the input image in the process of the proposed filter design and substitute the two 4f optical correlation system with a single 4f optical correlation system. The Performances of the proposed filter are compared with conventional CHF-SDF, POCHF-SDF filters through the computer simulation. The results of computer simulation show that the proposed filter has the rotation and scale-invariant correlation output and it has better performances than thoseof the conventioanl filters.

• 한국전기전자재료학회:학술대회논문집
• /
• 한국전기전자재료학회 2001년도 기술교육위원회 창립총회 및 학술대회 의료기기전시회
• /
• pp.34-37
• /
• 2001

This research suggested the new algorithm controlled by micro processor which is already stored by various PWM form of output voltage by using fundamental data of the carrier and signal. The determined PWM pattern is not concerned with the signal wave form and the new algorithm can obtain the desired pulse width by synchronous of carrier. The PWM wave can be controlled with real time by using extra hardware and digital software and to speed up program processing, the control signals to switch the power semi-conductor of three phase PWM inverter, simultaneously use the output signal by microprocessor and extra hardware, and control signal by software. In the end, this method was proved by applying to Three phase voltage source inverter.

• 대한전기학회논문지
• /
• 제40권7호
• /
• pp.696-705
• /
• 1991

This paper describes the procedures for pre-tuning and re-tuning the gains of PI controller based on output patterns -output error integral- of the unknown process which may not have any information, for example, system order, deadtime, time constant, etc. The key ideas of the proposed adaptive scheme are as follows. The scheme determines the initial gains by using ZNM (Ziegler-Nichols Method) with relay feedback, and then the adaptive algorithms by pattern recognition are introduced for re-runing the PI gains with on-line scheme whenever control conditions are changed. Because, among the various auto-tuning procedures, ANM with relay feedback has the difficulty in re-tuning with on-line and Bristol method has no comment on initial settings and has variables to pre-determine, which makes the algorithm comples, the proposed methods have the combined scheme with above two procedures to recover those problems. And this paper proposes a simple way to determine adaptive constant in Bristol method. To show the validity of the proposed method, an example is illustrated by computer simulation and a laboratory process, heat exchanger, is experimented.