• 예술인문사회 융합 멀티미디어 논문지
• /
• 제7권10호
• /
• pp.935-943
• /
• 2017

많은 양의 데이터 기반으로 학습하는 neural network 중 이미지 분류나 음성 인식 등에 사용되어 지고 있는 CNN(Convolution neural network)는 현재까지도 우수한 성능을 가진 구조로 계속적으로 발전되고 있다. 제한된 자원을 가진 임베디드 시스템에서 활용하기에는 많은 어려움이 있다. 그래서 미리 학습된 가중치를 사용하지만 여전히 한계점이 있기 때문에 이를 해결하기 위해 GPU의 범용 연산을 위해서 사용하는 GP-GPU(General-Purpose computing on Graphics Processing Units)를 활용하는 추세다. CNN은 단순하고 반복적인 연산을 수행하기 때문에 SIMT(Single Instruction Multiple Thread)기반의 GPGPU에서 스레드 할당과 활용 방법에 따라 연산 속도가 많이 달라진다. 스레드로 Convolution 연산과 Pooling 연산을 수행할 때 쉬어야 하는 스레드가 발생하는 데 이러한 문제를 해결하기 위해 남은 스레드가 다음 피쳐맵과 커널 계산에 활용되는 방법을 사용함으로써 연산 속도를 증가시켰다.

• 한국산학기술학회:학술대회논문집
• /
• 한국산학기술학회 2007년도 춘계학술발표논문집
• /
• pp.150-153
• /
• 2007

The intent of this paper is to describe a neural network structure called multi dynamic neural network(MDNN), and examine how it can be used in developing a learning scheme for computing robot inverse kinematic transformations. The architecture and learning algorithm of the proposed dynamic neural network structure, the MDNN, are described. Computer simulations are demonstrate the effectiveness of the proposed learning using the MDNN.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1993년도 한국자동제어학술회의논문집(국내학술편); Seoul National University, Seoul; 20-22 Oct. 1993
• /
• pp.842-846
• /
• 1993

Because of complexity, neural network is difficult to learn. So if internal representation[1] can be performed successfully, it is possible to use perceptron learning rule. As a result, learning is easier. Therefore the method of internal representations applied to the "XOR" problem, and the "spirals" problem. And then using the above results, the structure of neural network for computing is embodied.mputing is embodied.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1998년도 추계학술대회 논문집 학회본부 B
• /
• pp.533-536
• /
• 1998

The intent of this paper is to describe a neural network structure called dynamic neural processor(DNP), and examine how it can be used in developing a learning scheme for computing robot inverse kinematic transformations. The architecture and learning algorithm of the proposed dynamic neural network structure, the DNP, are described. Computer simulations are provided to demonstrate the effectiveness of the proposed learning using the DNP.

• 한국지능시스템학회:학술대회논문집
• /
• 한국퍼지및지능시스템학회 1998년도 The Third Asian Fuzzy Systems Symposium
• /
• pp.28-30
• /
• 1998

• 제어로봇시스템학회논문지
• /
• 제6권2호
• /
• pp.127-137
• /
• 2000

This paper presents a novel neural based controller which controls the water level of the nuclear power plant steam generator. The controller consists of a model reference feedback linearization controller and a PI controller for stabilizing the feedback linearization controller. The feedback linearization controller consists of a neural network model and an inversing module which uses the neural network model for computing the control input to the steam generator. We chose Piecewise Linearly Trained Network(PLTN) and Recurrent Neural Netwrok(RNN) for an approximator of the plant and used these approximators in calculating the input from the feedback linearization controller. Combining the above two controllers gives a result of better performance than the case which uses only a PI controller Each control result of PLTN and RNN is given.

• 한국지반공학회:학술대회논문집
• /
• 한국지반공학회 2000년도 가을 학술발표회 논문집
• /
• pp.673-680
• /
• 2000

Artificial neural networks are efficient computing techniques that are widely used to solve complex problems in many fields. In this study, a back-propagation neural network model for estimating a consolidation behavior of clay from soil parameter, site investigation data and the first settlement curve is proposed. The training and testing of the network were based on a database of 63 settlement curve from two different sites. Five different network models were used to study the ability of the neural network to predict the desired output to increasing degree of accuracy. The study showed that the neural network model predicted a consolidation behavior of clay reasonably well.

• Computers and Concrete
• /
• 제24권5호
• /
• pp.469-488
• /
• 2019

In this research study, the artificial neural networks approach is used to estimate the ultimate shear capacity of reinforced concrete beams with transverse reinforcement. More specifically, surrogate approaches, such as artificial neural network models, have been examined for predicting the shear capacity of concrete beams, based on experimental test results available in the pertinent literature. The comparison of the predicted values with the corresponding experimental ones, as well as with available formulas from previous research studies or code provisions highlight the ability of artificial neural networks to evaluate the shear capacity of reinforced concrete beams in a trustworthy and effective manner. Furthermore, for the first time, the (quantitative) values of weights for the proposed neural network model, are provided, so that the proposed model can be readily implemented in a spreadsheet and accessible to everyone interested in the procedure of simulation.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2003년도 학술회의 논문집 정보 및 제어부문 A
• /
• pp.171-174
• /
• 2003

In this paper, a Learning module for a reconfigurable neural network processor(ERNIE) was proposed for an On-chip learning. The existing reconfigurable neural network processor(ERNIE) has a much better performance than the software program but it doesn't support On-chip learning function. A learning module which is based on Back Propagation algorithm was designed for a help of this weak point. A pipeline structure let the learning module be able to update the weights rapidly and continuously. It was tested with five types of alphabet font to evaluate learning module. It compared with C programed neural network model on PC in calculation speed and correctness of recognition. As a result of this experiment, it can be found that the neural network processor(ERNIE) with learning module decrease the neural network training time efficiently at the same recognition rate compared with software computing based neural network model. This On-chip learning module showed that the reconfigurable neural network processor(ERNIE) could be a evolvable neural network processor which can fine the optimal configuration of network by itself.

• Structural Engineering and Mechanics
• /
• 제11권5호
• /
• pp.485-502
• /
• 2001

Modem soft computing methods, such as neural networks, evolutionary models and fuzzy logic, are mainly inspired by the problem solving strategies the biological systems use in nature. As such, the soft computing methods are fundamentally different from the conventional engineering problem solving methods, which are based on mathematics. In the author's opinion, these fundamental differences are the key to the full understanding of the soft computing methods and in the realization of their full potential in engineering applications. The main theme of this paper is to discuss the fundamental differences between the soft computing methods and the mathematically based conventional methods in engineering problems, and to explore the potential of soft computing methods in new ways of formulating and solving the otherwise intractable engineering problems. Inverse problems are identified as a class of particularly difficult engineering problems, and the special capabilities of the soft computing methods in inverse problems are discussed. Soft computing methods are especially suited for engineering design, which can be considered as a special class of inverse problems. Several examples from the research work of the author and his co-workers are presented and discussed to illustrate the main points raised in this paper.