• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1998.10a
• /
• pp.267-273
• /
• 1998

셀룰라 오토마타 신경망은 저자에 의하여 개발된 신경망으로써 주변의 셀과 국소적인 연결을 가지며 셀룰라 오토마타의 발생규칙에 따라 생성되는 신경망이다. 셀룰라 오토마타 신경망을 간단히 줄여서 ECANS라고 한다. 본 신경망은 카오스 뉴런 모델을 사용하며 뉴런사이의 연결강도는 흥분성 또는 억제성 결합을 갖는다. 신호의 전달방식은 펄스의 형태로서 뉴런이 발화하면 '1' 발화하지 않으면 '0'이 된다. 본 논문에서는 셀룰라 오토마타를 구성하는 요소별 특징을 살펴보고 주어진 문제에 적합한 셀룰라 오토마타 신경망을 얻어내기 위한 진화방법으로서 DNA 코딩방법을 제안한다. 제안한 방법의 유효성을 시뮬레이션을 통하여 검증한다.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.43 no.1 s.343
• /
• pp.17-26
• /
• 2006

The analog Viterbi decoder for the PR (1 2 2 1) which is used for BVD read channel is designed with circular analog parallel circuits. Since the inter symbol interference is serious problem in the high density magnetic storage device or DVD, the PRML technology is normally employed for the purpose of minimizing the decoding error. The feature of the PRML technology is with the multi-level coding according to the predetermined combining rule among the neighboring symbols and with the decoding according to the known combining rule. Though the conventional PRML technology is implemented with the digital circuits, the recent trend towards this end is with the utilization of the analog circuits due to the requirements of higher speed and lower power in the DVD read channel. In this study, the Viterbi decoder which is the bottleneck of the PRML implementation is designed with the analog parallel circuit structure. The designed Viterbi decoder for the PR (1 2 2 1) signal shows 3 times faster in the speed and 1/3 times less in the power consumption than thoseoftheconventionaldigitalcounterpart.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.8 no.6
• /
• pp.460-469
• /
• 2015

This study's aim is for reducing big data transfer time of IoT healthcare devices by modulating digital bits into Manchester code including zero-voltage idle as information for secondary compressed binary cluster's compartment after two step compression of compressing binary data into primary and secondary binary compressed clusters for each binary clusters having compression benefit of 1 bit or 2 bits. Also this study proposed that as department information of compressed binary clusters, inserting idle signal into Manchester code will have benefit of reducing transfer time in case of compressing binary cluster into secondary compressed binary cluster by 2 bits, because in spite of cost of 1 clock idle, another 1 bit benefit can play a role of reducing 1 clock transfer time. Idle signal is also never consecutive because the signal is for compartment information between two adjacent secondary compressed binary cluster. Voltage transition on basic rule of Manchester code is remaining while inserting idle signal, so DC balance can be guaranteed. This study's simulation result said that even compressed binary data by another compression algorithms could be transferred faster by as much as about 12.6 percents if using this method.

• Journal of Korean Library and Information Science Society
• /
• v.41 no.2
• /
• pp.251-272
• /
• 2010

Resource Description and Access(RDA) was published in 2008, reflecting IFLA IME ICC's International Cataloguing Principles(ICP), conceptual model for cataloging of Functional Requirement for Bibliographic Records(FRBR), and conceptual model for authority data of Functional Requirement for Authority Data(FRAD). Therefore, new tags and fields have been added to MARC21 format, and researches about the prototype have been studied to implement FRBR as envisioned by the new RDA standard. According to the innovative international cataloging environments, KCR and KORMARC must be revised to accommodate RDA description. This study was to grasp occurrence rate of work and expression element of FRBR and RDA among MARC field of Hansung University library and to analyze limitation of MARC data for RDA. Based on this limitations, cataloging rules and tags and fields of KORMARC were suggested to accommodate RDA in national cataloging environment. This study reviewed cataloging rules such like representation, main entry and rule of three, preferred title, types of content, media and carrier. This study is to suggest the consideration of KCR and KORMARC for RDA in Korea.

• KIISE Transactions on Computing Practices
• /
• v.23 no.9
• /
• pp.542-549
• /
• 2017

Sequence-to-sequence models and similar pointer networks suffer from performance degradation when an input is composed of multiple sentences or when the length of the input sentence is long. To solve this problem, this paper proposes a hierarchical pointer network model that uses both the word level and sentence level information to encode input sequences composed of several sentences at the word level and sentence level. We propose a hierarchical pointer network based coreference resolution that performs a coreference resolution for all mentions. The experimental results show that the proposed model has a precision of 87.07%, recall of 65.39% and CoNLL F1 74.61%, which is an improvement of 21.83% compared to an existing rule-based model.

• Journal of the Korean Institute of Telematics and Electronics D
• /
• v.35D no.10
• /
• pp.107-115
• /
• 1998

A multiplier bit-pair recoding technique derived from Booth algorithm is used as an effective method that can carry out a fast binary multiplication regardless of a sign of both multiplicand and multiplier. In this paper, we propose an implementation of an optical high-speed multiplier which consists of a symbolic substitution adder and an optical multiplication algorithm, which transforms and enhances the multiplier bit-pair recoding algorithm to be fit for optical characteristics. Specially, a symbolic substitution addition rules are coded with a dual-rail logic, and so the complement of the logic of the symbolic substitution adder is easily obtained with a shift operation because it is always present. We also construct the symbolic substitution system which makes superposition image by superimposing two shifted images in a serial connection and recognizes a reference image by feeding this superimposed image to a mask. Thus, the optical multiplier, which is compared with a typical system, is implemented to the smaller system by reducing the number of optical passive elements and the size of this system.

• Proceedings of the Korean Information Science Society Conference
• /
• 2000.10b
• /
• pp.197-199
• /
• 2000

연속음성인식을 수행하기 위해서는 발음사전과 언어모델이 필요하다. 이 둘 사이에는 디코딩 단위가 일치하여야 하므로 발음사전 구축시 디코딩 단위로 표제어 단위를 선정하며 표제어 사이의 음운변화 현상을 반영한 발음사전을 구축하여야 한다. 한국어에 부합하는 음운변화현상을 분석하여 학습용 자동 발음열을 생성하고, 이를 통하여 발음사전을 구축한다. 전처리 단계로 기호, 단위, 숫자 등 전처리 과정 및 형태소 분석 과정을 수행하며, 디코딩 단위인 의사 형태소 단위를 생성하기 위해 규칙을 이용한 태깅 과정을 거친다. 이를 통해 나온 결과를 발음열 생성기 입력으로 하며, 결과는 학습용 발음열 또는 발음사전 구성을 위한 형태로 출력한다. 표제어간 음운변화 현상이 반영된 상태의 표제어 단위이므로 실제 음운변화가 반영되지 않은 상태의 표제어와는 그 형태가 상이하다. 이는 연속 발음시 생기는 현상으로 실제 인식에는 이 음운변화 현상이 반영된 사전이 필요하게 된다. 생성된 발음사전의 효용성을 확인하기 위해 다음과 같은 실험을 통해 성능을 평가하였다. 음향학습을 위하여 PBS(Phonetically Balanced Sentence) 낭독체 17200문장을 녹음하고 그 전사파일을 사용하여 학습을 수행하였고, 발음사전의 평가를 위하여 이 중 각각 3100문장을 사용하여 다음과 같은 실험을 수행하였다. 형태소 태그정보를 이용하여 표제어간 음운변화 현상을 반영한 최적의 발음사전과 다중 발음사전, 언어학적 기준에 의한 수작업으로 생성한 표준 발음사전, 그리고 표제어간의 음운변화 현상을 고려하지 않고 독립된 단어로 생성한 발음사전과의 비교 실험을 수행하였다. 실험결과 표제어간 음운변화 현상을 반영하지 않은 경우 단어 인식률이 43.21%인 반면 표제어간 음운변화 현상을 반영한 1-Best 사전의 경우 48.99%, Multi 사전의 경우 50.19%로 인식률이 5~6%정도 향상되었음을 볼 수 있었고, 수작업에 의한 표준발음사전의 단어 인식률 45.90% 보다도 약 3~4% 좋은 성능을 보였다.

• Smart Media Journal
• /
• v.7 no.4
• /
• pp.17-23
• /
• 2018

We proposed a LSTM-based RNN model that can effectively perform the automatic spacing characteristics. For those long or noisy sentences which are known to be difficult to handle within Neural Network Learning, we defined a proper input data format and decoding data format, and added dropout, bidirectional multi-layer LSTM, layer normalization, and attention mechanism to improve the performance. Despite of the fact that Sejong corpus contains some spacing errors, a noise-robust learning model developed in this study with no overfitting through a dropout method helped training and returned meaningful results of Korean word spacing and its patterns. The experimental results showed that the performance of LSTM sequence-to-sequence model is 0.94 in F1-measure, which is better than the rule-based deep-learning method of GRU-CRF.

• Journal of Korean Library and Information Science Society
• /
• v.52 no.1
• /
• pp.109-128
• /
• 2021

This paper is to explore considerations in authority control to accommodate LRM nomen entities through the literature reviews, the analysis of RDA rules, and the opinion survey of domestic catalog experts. As a result, for authority control, considerations were proposed in the aspect of nomen's attribute elements, catalog description, and MARC authority format. First, it is necessary to describe in as much detail as possible the category, the scheme, intended audience, the context of use, the reference source, the language, the script, the script conversion as the attributes of the nomen with the status of identification, note, and indifferentiated name indicators added in RDA. Second, the description method of attribute elements and relational elements of nomen can be unstructured, structured, identifier, and IRI as suggested in RDA, and vocabulary encoding scheme (VES) and string encoding scheme (SES) should be written for structured description, Also, cataloging rules for structuring authorized access points and preferred names/title should be established. Third, an additional expansion plan based on Maxwell's expansion (draft) was proposed in order to prepare the MARC 21 authority format to reflect the LRM nomen. (1) The attribute must be described in 4XX and 5XX so that the attribute can be entered for each nomen, and the attributes of the nomen to be described in 1XX, 5XX and 4XX are presented separately. (2) In order to describe the nomen category, language, script, script conversion, context of use, and date of usage as a nomen attribute, field and subfield in MARC 21 must be added. Accordingly, it was proposed to expand the subfield of 368, 381, and 377, and to add fields to describe the context of use and date of usage. The considerations in authority control for the LRM nomen proposed in this paper will be the basis for establishing an authority control plan that reflects LRM in Korea.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.13 no.2
• /
• pp.181-189
• /
• 2013

In this paper, we introduce the fuzzy neural network based on the individual input space to design the pattern recognizer. The proposed networks configure the network by individually dividing each input space. The premise part of the networks is independently composed of the fuzzy partition of individual input spaces and the consequence part of the networks is represented by polynomial functions. The learning of fuzzy neural networks is realized by adjusting connection weights of the neurons in the consequent part of the fuzzy rules and it follows a back-propagation algorithm. In addition, in order to optimize the parameters of the proposed network, we use real-coded genetic algorithms. Finally, we design the optimized pattern recognizer using the experimental data for pattern recognition.