• 한국컴퓨터정보학회:학술대회논문집
• /
• 한국컴퓨터정보학회 2016년도 제53차 동계학술대회논문집 24권1호
• /
• pp.71-72
• /
• 2016

베이지안 알고리즘은 추론 분야에서 오랜 기간 사용되어 왔다. 하지만 기본적인 베이지안 네트워크 이론만으로는 다양한 도메인에 적합한 추론 기능을 제공할 수 없기 때문에, 도메인의 특성에 맞는 알고리즘이 적용된 다양한 추론 기법들이 연구되어왔다. 본 논문에서는 실시간 지능화 서비스를 위하여 특정 도메인 영역에 대하여 자동으로 적합한 베이지안 네트워크 알고리즘을 선별하는 기법을 제안한며, 해당 기법의 적합도를 평가하기 위해서 수학적인 모델링과 추론 알고리즘 선택 기법에 대해 서술한다.

• ETRI Journal
• /
• 제43권4호
• /
• pp.684-693
• /
• 2021

This paper presents a low-cost two-stage approximate multiplier for bfloat16 (brain floating-point) data processing. For cost-efficient approximate multiplication, the first stage implements Mitchell's algorithm that performs the approximate multiplication using only two adders. The second stage adopts the exact multiplication to compensate for the error from the first stage by multiplying error terms and adding its truncated result to the final output. In our design, the low-cost multiplications in both stages can reduce hardware costs significantly and provide low relative errors by compensating for the error from the first stage. We apply our approximate multiplier to the convolutional neural network (CNN) inferences, which shows small accuracy drops with well-known pre-trained models for the ImageNet database. Therefore, our design allows low-cost CNN inference systems with high test accuracy.

• 한국정보처리학회논문지
• /
• 제4권4호
• /
• pp.1025-1034
• /
• 1997

레졸루션(resolution)에 근거한 자동화된 추론 방법은 실제로 결론 도출에 필요 하지 않은 중간 정보를 생산하고, 또한 이런 정보가 또 다른 불 필요 정보를 생산함으로써, 컴퓨터 주기억 공간 잠식이 문제점으로 나타난다. 주어진 문제의 상황을 설명하는 모든 입력절로부터, 실제로 결론의 도출에 참여할 가능성이 낮은 입력절을 추론에의 참여를 제한하여, 소모하는 기억공간과 추론 시간을 감소시키는 조절 전략을 제안한다. 주어진 입력절을 분석하여, 실제로 추론에 참여하는 정도의 우선 순위를 결정하는 도구로서 퍼지 관계 논리곱을 이용한다.

• 한국항행학회논문지
• /
• 제12권4호
• /
• pp.378-383
• /
• 2008

본 논문에서는 DCT 변환과 퍼지추론을 이용하여 영상을 분석하는 방법을 제안 한 바, 병해충 과실 등의 특성을 분석 할 수 있는 퍼지추론 알고리즘과 변환계수에 시각특성파라미터를 접목하는 방법에 중점을 두었다. 전처리 과정에서 이산코사인 변환계수로부터 엔트로피와 텍스처 등의 시각특징 파라미터들을 구하였고, 이 변수들을 이용하여 퍼지 추론의 입력 변수를 생성 하였다. 맘다니 연산자와 ${\alpha}$-cut 함수를 적용하여 영상 분석을 실험한 결과, 제안한 방법의 응용가능성을 입증하였다.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2004년도 학술대회 논문집 정보 및 제어부문
• /
• pp.340-342
• /
• 2004

In this paper, we introduce a new category of fuzzy inference systems baled on information granulation to carry out the model identification of complex and nonlinear systems. Informal speaking, information granules are viewed as linked collections of objects(data, in particular) drawn together by the criteria of proximity, similarity, or functionality. Granulation of information with the aid of Hard C-Means(HCM) clustering algorithm help determine the initial parameters of fuzzy model such as the initial apexes of the membership functions and the initial values of polyminial functions being used in the premise and consequence part of the fuzzy rules. And the initial parameters are tuned effectively with the aid of the genetic algorithms(GAs) and the least square method. The proposed model is contrasted with the performance of the conventional fuzzy models in the literature.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2004년도 학술대회 논문집 정보 및 제어부문
• /
• pp.343-345
• /
• 2004

We propose a new category of fuzzy set-based fuzzy inference systems based on data granulation related to fuzzy space division for each variables. Data granules are viewed as linked collections of objects(data, in particular) drawn together by the criteria of proximity, similarity, or functionality. Granulation of data with the aid of Hard C-Means(HCM) clustering algorithm help determine the initial parameters of fuzzy model such as the initial apexes of the membership functions and the initial values of polyminial functions being used in the premise and consequence part of the fuzzy rules. And the initial parameters are tuned effectively with the aid of the genetic algorithms(GAs) and the least square method. Numerical example is included to evaluate the performance of the proposed model.

• 전자공학회논문지B
• /
• 제31B권1호
• /
• pp.99-106
• /
• 1994

In this paper, we develop an algorithm which recognizes Ilpa-style shorthand characters by on-line. It discriminates the structure of characters using coordinates which are measured by tablet board, then it outputs the recognized characters using the fuzzy inference rules. Shorthand characters have several forms, in which an initial or a middle sound depends on angle and length while a last sound is treated as a hook. We apply fuzzy inference rules to the discrimination of the length, the angle, the curve, and the straight line. We also built up a set of standard character codes in order to reduce the processing time.

• 전기학회논문지
• /
• 제57권3호
• /
• pp.473-479
• /
• 2008

Generally, HDDs are widely used as data storage device in office, home and mobile machineries. So, it is used for various applications or tasks, such as file copy, file download, music and movie play etc., in various environment. In spite of this kind of varieties in tasks and environment in which HDDs perform, most commercial HDDs hardly control its operations adaptively to these varieties. Thus, it is preferred to optimize the performance and energy consumption of HDDs according to the task and/or the environment. So, this paper proposes a new fuzzy inference algorithm which adaptively controls HDDs operations and may also easily be implemented as the firmware of HDDs and run in the restricted environment such as embedded systems.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• 제4권1호
• /
• pp.50-55
• /
• 2004

We propose an intgrated Gill environment system having only integer fuzzy operations in the consequent part and the defuzzification stage. In this paper, we also propose an integrated Gill environment system with 4 parallel fuzzy processing units to be operated in parallel on the classification of the sensed image data. In this, we solve the problems of taking longer times as the fuzzy real computations of [0, 1] by using the integer pixel conversion algorithm to convert lines of each fuzzy linguistic term to the closest integer pixels. This procedure is performed automatically in the GUI application program. As a Gill environment, PCI transmission, image data pre-processing, integer pixel mapping and fuzzy membership tuning are considered. This system can be operated in parallel manner for MIMO or MISO systems.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2006년도 제37회 하계학술대회 논문집 D
• /
• pp.2077-2078
• /
• 2006

This paper describes the development or a closed-loop Strapdown Attitude Reference System (SARS) algorithm integrated filtering estimator for determining attitude reference for railway and aviation system using fuzzy inference. The SARS consists of 3 single-axis rate gyms in conjunction with 2 single-axis accelerometers. For optimal values of fuzzy systems, we utilize on-line scheduling method for initial values and then use genetic algorithms for fine tuning. Implementation using experimental test data of unmanned aerial vehicle has been performed in order to verify the estimation. The proposed fuzzy inference based SARS demonstrate that more accurate performance can be achieved in comparison with conventional one. The estimation results were compared with the on-board vertical gyro as the reference standard.