• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.33B no.9
• /
• pp.117-123
• /
• 1996

In this paper, we present a new learning method for the fuzzy ARTMAP which is effective for the noisy input patterns. Conventional fuzzy ARTMAP employs only fuzzy AND operation between input vector and weight vector in learning both top-down and bottom-up weight vectors. This fuzzy AND operation causes excessive update of the weight vector in the noisy input environment. As a result, the number of spurious categories are increased and the recognition ratio is reduced. To solve these problems, we propose a new method in updating the weight vectors: the top-down weight vectors of the fuzzy ART system are updated using weighted average of the input vector and the weight vector itself, and the bottom-up weight vectors are updated using fuzzy AND operation between the updated top-down weitht vector and bottom-up weight vector itself. The weighted average prevents the excessive update of the weight vectors and the fuzzy AND operation renders the learning fast and stble. Simulation results show that the proposed method reduces the generation of spurious categories and increases the recognition ratio in the noisy input environment.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.31B no.9
• /
• pp.133-141
• /
• 1994

Conventional competitive learning algorithms compute the Euclidien distance to determine the winner neuron out of all predetermined output neurons. In such cases, there is a drawback that the performence of the learning algorithm depends on the initial reference(=weight) vectors. In this paper, we propose a new competitive learning algorithm that dynamically generates output neurons. The proposed method generates output neurons by dynamically changing the class thresholds for all output neurons. We compute the similarity between the input vector and the reference vector of each output neuron generated. If the two are similar, the reference vector is adjusted to make it still more like the input vector. Otherwise, the input vector is designated as the reference vector of a new outputneuron. Since the reference vectors of output neurons are dynamically assigned according to input pattern distribution, the proposed method gets around the phenomenon that learning is early determined due to redundant output neurons. Experiments using speech data have shown the proposed method to be superior to existint methods.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.10
• /
• pp.1358-1363
• /
• 1995

This paper presents a new learning method for a noise tolerant Fuzzy ART. In the conventional Fuzzy ART, the top-down and bottom-up weight vectors have the same value. They are updated by a fuzzy AND operation between the input vector and the current value of the top-down or bottom- up weight vectors. However, it can not prevent the abrupt change of the weight vector and can not achieve good performance for a noisy input vector. To solve the problems, we updated using the weighted sum of the input vector and the current value of the top-down vector. To achieve stability, the bottom-up weight vector is updated using the fuzzy AND operation between the newly learned top-down vector and the current value of the bottom-up vector. Computer simulations show that the proposed method prominently resolves the category proliferation problem without increasing the training epoch for stabilization in noisy environments.

• 전기의세계
• /
• v.23 no.1
• /
• pp.69-72
• /
• 1974

This paper deals with an analytical design of a feedback regulator with vector input is discrete linear time-invariant systems. We have derived some relations such that the eigenvalues of a system plant with vector input under the time-optimal control strategy can be arbitrarily changed by the characteristics of the minor loop compensator which is indroduced in the feedback path.

• Proceedings of the IEEK Conference
• /
• 2005.11a
• /
• pp.655-658
• /
• 2005

Due to reduced device sizes and threshold voltages, leakage current becomes an important issue in CMOS design. In a CMOS combinational logic circuit, the leakage current in the standby state depends on the state of the inputs and thus can be minimized by applying an optimal input when the circuit is idling. In this paper, we present a New Input Vector Control algorithm, called Leakage Minimization by Input vector Control (LMIC) for minimal leakage power. This algorithm finds the minimal leakage vector and reduces leakage current up to 22.% on the average, for TSMC 0.18um process parameters. Minimal leakage vectors are very useful in reducing leakage currents in standby mode of operation.

• Proceedings of the Optical Society of Korea Conference
• /
• 1990.02a
• /
• pp.102-111
• /
• 1990

In this thesis, an optical bidirectional inner-product associative memory model using liquid crystal television is proposed and analyzed theoretically and realized experimentally. The LCTV is used as a SLM(spatial light modulator), which is more practical than conventional SLMs, to produce image vector in terms of computer and CCD camera. Memory and input vectors are recorded into each LCTV through the video input connectors of it by using the image board. Two multi-focus hololenses are constructed in order to perform optical inner-product process. In forward process, the analog values of inner-products are measured by photodetectors and are converted to digital values which are enable to control the weighting values of the stored vectors by changing the gray levels of the pixels of the LCTV. In backward process, changed stored vectors are used to produce output image vector which is used again for input vector after thresholding. After some iterations, one of the stored vectors is retrieved which is most similar to input vector in other words, has the nearest hamming distance. The experimental results show that the proposed inner-product associative memory model can be realized optically and coincide well with the computer simulation.

• Proceedings of the KSME Conference
• /
• 2004.11a
• /
• pp.817-822
• /
• 2004

This paper provides a global path planning method using self-organizing feature map which is a method among a number of neural network. The self-organizing feature map uses a randomized small valued initial weight vectors, selects the neuron whose weight vector best matches input as the winning neuron, and trains the weight vectors such that neurons within the activity bubble are move toward the input vector. On the other hand, the modified method in this research uses a predetermined initial weight vectors, gives the systematic input vector whose position best matches obstacles, and trains the weight vectors such that neurons within the activity bubble are move toward the input vector. According to simulation results one can conclude that the modified neural network is useful tool for the global path planning problem of a mobile robot.

• Journal of the Korean Society for Precision Engineering
• /
• v.23 no.3 s.180
• /
• pp.94-101
• /
• 2006

A global path planning method using self-organizing feature map which is a method among a number of neural network is presented. The self-organizing feature map uses a randomized small valued initial weight vectors, selects the neuron whose weight vector best matches input as the winning neuron, and trains the weight vectors such that neurons within the activity bubble are moved toward the input vector On the other hand, the modified method in this research uses a predetermined initial weight vectors of 1-dimensional string and 2-dimensional mesh, gives the systematic input vector whose position best matches obstacles, and trains the weight vectors such that neurons within the activity bubble are moved toward the input vector. According to simulation results one can conclude that the modified neural network is useful tool for the global path planning problem of a mobile robot.

• Phonetics and Speech Sciences
• /
• v.7 no.4
• /
• pp.3-8
• /
• 2015

In this paper, we investigate an input dimension reduction method using continuous word vector in deep neural network language model. In the proposed method, continuous word vectors were generated by using Google's Word2Vec from a large training corpus to satisfy distributional hypothesis. 1-of-${\left|V\right|}$ coding discrete word vectors were replaced with their corresponding continuous word vectors. In our implementation, the input dimension was successfully reduced from 20,000 to 600 when a tri-gram language model is used with a vocabulary of 20,000 words. The total amount of time in training was reduced from 30 days to 14 days for Wall Street Journal training corpus (corpus length: 37M words).

• Journal of the Korean Society for Precision Engineering
• /
• v.25 no.4
• /
• pp.69-76
• /
• 2008

The self-organizing feature map(SOFM) among a number of neural network uses a randomized small valued initial weight vectors, selects the neuron whose weight vector best matches input as the winning neuron, and trains the weight vectors such that neurons within the activity bubble are moved toward the input vector. On the other hand, the modified method in this research uses a predetermined initial weight vectors of the 1-dimensional string, gives the systematic input vector whose position best matches obstacles, and trains the weight vectors such that neurons within the activity bubble are move toward the opposite direction of input vector. According to simulation results one can conclude that the method using string and the modified neural network is useful tool to mobile robot for the global path planning.