• 정보처리학회논문지B
• /
• 제14B권4호
• /
• pp.287-294
• /
• 2007

실세계의 문제에서 많은 기계학습의 알고리즘들은 데이터의 클래스 불균형 문제에 어려움을 겪는다. 이러한 클래스 불균형 문제를 해결하기 위하여 데이터의 비율을 변경하거나 좀 더 나은 샘플링 전략으로 극복하려는 연구들이 제안되었다. 그러나 데이터의 비율을 변경하는 연구에서는 전체 데이터 분포의 특성을 고려하지 못하고, 샘플링 전략을 제안하는 연구에서는 여러 가지 제한 조건을 고려해야만 한다. 본 논문에서는 위의 두가지 방법의 장점을 모두 포함하는 개선된 집중 샘플링 방법을 제안한다. 제안된 방법에서는 클래스 불균형 문제를 해결하기 위해 학습에 유용한 데이터들을 샘플링하는데 스코어링에 기반한 데이터 분할 방법을 이용한다. 즉, 입력 데이터들에 대해 SOM(Self Organizing Map)의 학습 결과로 얻은 BMU(Best Matching Unit)와의 거리를 계산하고, 이 거리론 스코어라 한다. 측정된 스코어는 오름차순으로 정렬되며, 이 과정에서 입력 데이터의 분포가 재 표현되고, 재 표현된 분포는 전체 데이터의 특성을 대표하게 된다. 그 결과로 얻은 데이터들 중에서 유용하지 못한 데이터들에 대해 제거하는 과정을 수행하여 새로운 학습 데이터 셋을 얻는다. 새로운 학습 데이터 생성 과정에서는 재 표현된 분포의 결과를 두 구간(upper, lower)으로 분할하는데, 두 추간 사이의 데이터들은 유용하지 못한 패턴들로 간주되어 학습에 이용되지 않는다. 본 논문에서 제안한 방법은 클래스 불균형의 비율 감수 훈련 데이터의 크기 감소, 과적합의 방지 등 몇 가지 장점을 보인다. 제안한 방법으로 샘플링된 데이터에 kNN 을 적용하여, 분류 실험한 결과 심한 불균형이 있는 ecoli 데이터의 분류 성능이 최대 2.27배 향상되었다.

• 대한전자공학회논문지SD
• /
• 제43권11호
• /
• pp.37-47
• /
• 2006

본 논문에서는 Digital Video Broadcasting (DVB), Digital Audio Broadcasting (DAB) 및 Digital Multimedia Broadcasting (DMB) 등과 같이 저전압, 저전력 및 소면적을 동시에 요구하는 고성능 무선 통신 시스템을 위한 10b 25MS/s $0.8mm^2$ 4.8mW 0.13um CMOS A/D 변환기 (ADC)를 제안한다. 제안하는 ADC는 요구되는 해상도 및 속도 사양을 만족시키면서 동시에 면적 및 전력 소모를 최소화하기 위해 2단 파이프라인 구조를 사용하였으며, 스위치 기반의 바이어스 전력 최소화 기법(switched-bias power reduction technique)을 적용하여 전체 전력 소모를 최소화하였다. 입력단 샘플-앤-홀드 증폭기는 낮은 문턱전압을 가진 트랜지스터로 구성된 CMOS 샘플링 스위치를 사용하여 10비트 이상의 해상도를 유지하면서, Nyquist rate의 4배 이상인 60MHz의 높은 입력 신호 대역폭을 얻었으며, 전력소모를 최소화하기 위해 1단 증폭기를 사용하였다. 또한, Multiplying D/A 변환기의 커패시터 열에는 소자 부정합에 의한 영향을 최소화하기 위해서 인접신호에 덜 민감한 3차원 완전 대칭 구조의 커패시터 레이아웃 기법을 제안하며, 기준 전류 및 전압 발생기는 온-칩으로 집적하여 잡음을 최소화하면서 필요시 선택적으로 다른 크기의 기준 전압을 외부에서 인가할 수 있도록 설계하였다. 또한, 다운 샘플링 클록 신호를 사용하여 바이어스 전류를 제어함으로써 10비트의 해상도에서 응용 분야에 따라서 25MS/s 뿐만 아니라 10MS/s의 동작 속도에서 더 낮은 전력 사용이 가능하도록 하였다. 제안하는 시제품 ADC는 0.13um 1P8M CMOS 공정으로 제작되었으며 측정된 최대 DNL 및 INL은 각각 0.42LSB 및 0.91LSB 수준을 보인다. 또한, 25MS/s 및 10MS/s의 동작 속도에서 최대 SNDR 및 SFDR이 각각 56dB, 65dB이고, 전력 소모는 1.2V 전원 전압에서 각각 4.8mW, 2.4mW이며 제작된 ADC의 칩 면적은 $0.8mm^2$이다.

• 벤처창업연구
• /
• 제16권2호
• /
• pp.35-49
• /
• 2021

본 연구의 목적은 창업 인프라 기관인 N사의 파트너스퀘어(광주)를 이용한 경험이 있는 청년자영업자를 대상으로 온라인쇼핑몰에서 창업하는 과정을 근거이론(Grounded Theory) 방법을 활용해 규명해 보는 데 있다. 본 연구는 심층 면접 조사 자료를 활용했는데, 연구 참여자의 선정에 있어 이론적 표본 추출 (Theoretical Sampling)방식을 사용했으며, 스트라우스와 코빈(Strauss & Corbin)이 제시한 개방코딩, 축 코딩, 선택코딩 순으로 진행한 후 패러다임모형으로 분석했다. 주요 연구 결과는 다음과 같다. 첫째, 부모는 온라인쇼핑몰에 대해 잘 알지 못하거나 부정적이라도 자녀의 창업에 대해서 긍정적이면 자녀가 온라인쇼핑몰에서 창업하는 것으로 나타났다. 그러나 부모가 온라인쇼핑몰에 대해서 부정적이고, 자녀의 창업에도 부정적이면 자녀는 온라인쇼핑몰에서 창업하지 못하는 것으로 나타났다. 둘째, 청년층의 온라인쇼핑몰에서 창업과 성과 창출 조건으로 온라인쇼핑몰 활용능력이 중요한 것으로 나타났는데, 특히 파트너스퀘어 (광주)는 온라인쇼핑몰 활용능력을 증대시켜, 청년층의 온라인쇼핑몰에서 창업에 긍정적 영향을 주는 것으로 나타났다. 셋째, 청년층은 온라인쇼핑몰에서 창업 이후 단순한 매출액 증가 이외에 자존감 상승, 기회발견, 창의성 역량이 강화되는 것으로 나타났다.

• Journal of Microbiology and Biotechnology
• /
• 제22권8호
• /
• pp.1148-1154
• /
• 2012

A novel two-phase anaerobic treatment technology was developed to treat high-concentration organic cassava bioethanol wastewater. The start-up process and contribution of organics (COD, total nitrogen, and $NH_4^+$-N) removal in spatial succession of the whole process and spatial microbial diversity changing when sampling were analyzed. The results of the start-up phase showed that the organic loading rate could reach up to $10kg\;COD/m^3d$, with the COD removal rate remaining over 90% after 25 days. The sample results indicated that the contribution of COD removal in the pre-anaerobic and anaerobic phases was 40% and 60%, respectively, with the highest efficiency of 98.5%; TN and $NH_4^+$-N had decreased to 0.05 g/l and 0.90 g/l, respectively, and the mineralization rate of total nitrogen was 94.8%, 76.56% of which was attributed to the anaerobic part. The microbial diversity changed remarkably among different sample points depending on the physiological characteristics of identified strains. Moraxellaceae, Planococcaceae, and Prevotellaceae were dominant in the pre-anaerobic phase and Bacteroidetes, Campylobacterales, Acinetobacter, Lactobacillus, Clostridium, and Bacillus for the anaerobic phase. Methanosarcinaceae and Methanosaeta were the two main phylotypes in the anaerobic reactor.

• 대한조선학회논문집
• /
• 제60권3호
• /
• pp.202-211
• /
• 2023

The profile shops in shipyards produce section steels required for block production of ships. Due to the limitations of shipyard's production capacity, a considerable amount of work is already outsourced. In addition, the need to improve the productivity of the profile shops is growing because the production volume is expected to increase due to the recent boom in the shipbuilding industry. In this study, a scheduling optimization was conducted for a parallel welding line of the profile process, with the aim of minimizing tardiness and the number of set-up changes as objective functions to achieve productivity improvements. In particular, this study applied a dynamic scheduling method to determine the job sequence considering variability of processing time. A Markov decision process model was proposed for the job sequence problem, considering the trade-off relationship between two objective functions. Deep reinforcement learning was also used to learn the optimal scheduling policy. The developed algorithm was evaluated by comparing its performance with priority rules (SSPT, ATCS, MDD, COVERT rule) in test scenarios constructed by the sampling data. As a result, the proposed scheduling algorithms outperformed than the priority rules in terms of set-up ratio, tardiness, and makespan.

• 한국환경보건학회지
• /
• 제31권3호
• /
• pp.215-220
• /
• 2005

The atmospheric transportation and dispersion processes of air pollutants are important issues in dealing with air pollution problems. Air pollutants originated from biological and anthropogenic activities are not only limited to the local emission sources, but could also be transported and dispersed to other regions by synoptic weather systems. Besides, the complexity of topography of central Taiwan helps accumulating air pollutants to promote high-concentration episodes. The techniques of tethersonding were applied to monitor the vertical profiles of winds, air temperatures and humidity, as well as to collect air samples, to be analyzed for pollutants $(O_3,\;NO_2,\;NO\;and\; NMHC)$ from the ground up to 1000m. A time period of about one week, 19-26 October 2002, was chosen as the sampling period due to the high frequency of episode occurrence in autumn based on the past records. Associating with the analysis of weather patterns, the atmospheric characteristics over high-concentration areas can be resolved in more detail. The result of the tethersonding studies showed that weak northerly sea breeze (with thickness about 300m) with low wind speed (about 1 to 2 m/sec) could help develop high ozone concentrations in the down-wind areas. It is also important to have a built-up aloft of precursors and ozone to develop high concentration on the previous day.

• 한국환경보건학회:학술대회논문집
• /
• 한국환경보건학회 2005년도 국제학술대회
• /
• pp.184-210
• /
• 2005

The atmospheric transportation and dispersion processes of air pollutants are important issues in dealing with air pollution problems. Air pollutants originated from biological and anthropogenic activities are not only limited to the local emission sources, but could also be transported and dispersed to other regions by synoptic weather systems. Besides, the complexity of topography of central Taiwan helps accumulating air pollutants to promote high-concentration episodes. The techniques of tethersonding were applied to monitor the vertical profiles of winds, air temperatures and humidity, as well as to collect air samples, to be analyzed for pollutants ($O_3,\;NO_2$, NO and NMHC) from the ground up to 1000 m. A time period of about one week, 19 -26 October 2002, was chosen as the sampling period due to the high frequency of episode occurrence in autumn based on the past records. Associating with the analysis of weather patterns, the atmospheric characteristics over high-concentration areas can be resolved in more detail. The result of the tethersonding studies showed that weak northerly sea breeze (with thickness about 300 m) with now wind speed (about 1 to 2 m/sec) could help develop high ozone concentrations in the down-wind areas. It is also important to have a built-up aloft of precursors and ozone to develop high concentration on the previous day.

• 대한의용생체공학회:의공학회지
• /
• 제16권1호
• /
• pp.107-113
• /
• 1995

It is well known that multipoint and computerized intraoperative mapping systems improve the results of surgery for Wolff-Parkinson-White syndrome and show tremendous potential for opening an entirely new era of surgical intervention for the more common and lethal types of supraventricular tachyarrhythmias such as atrial flutter and atrial fibrillation. In addition, the ability to map and ablate the sometimes fleeting automatic atrial tachycardia is greatly enhanced by computerized mapping systems. In this study, we have developed 64 channel computerized data analysis system using microcomputer (Macintosh ${II}_{x}$) for basic research of electrophysiology and electrical propagation. The bipolar electrogram information is acquired from 64 cardiac sites simultaneously at a sampling rate of 1 ksampls/sec with continuous and total data storage of up to 30 seconds. When the reference electrogram is selected and reference point is picked up, delay time from the reference point is displayed on two dimensional diagram of the heart. System design permits easy expansion to almost 256 simultaneous sites. this system is expected to enable us to study pathophysiology of cardiac arrhythmia and to improve the result of diagnosis and surgical treatment for cardiac arrhythmia.

• 대한기계학회논문집A
• /
• 제24권9호
• /
• pp.2361-2370
• /
• 2000

The introduction of inexpensive cylinder pressure sensors provides new opportunities for precise engine control. This paper presents a control strategy of spark advance based upon cylinder pressure of spark ignition engines. A location of peak pressure(LPP) is the major parameter for controlling the spark timing, and also the UP is estimated, using a multi-layer feedforward neural network, which needs only five pressure sensor output voltage samples at -40˚, -20˚, 0˚, 20˚, 40˚ after top dead center. The neural network plays an important role in mitigating the A/D conversion load of an electronic engine controller by increasing the sampling interval from 10 crank angle(CA) to 20˚ CA. A proposed control algorithm does not need a sensor calibration and pegging(bias calculation) procedure because the neural network estimates the UP from the raw sensor output voltage. The estimated LPP can be regarded as a good index for combustion phasing, and can also be used as an MBT control parameter. The feasibility of this methodology is closely examined through steady and transient engine operations to control individual cylinder spark advance. The experimental results have revealed a favorable agreement of individual cylinder optimal combustion phasing.

• 한국의류학회지
• /
• 제32권7호
• /
• pp.1065-1074
• /
• 2008

The purpose of this paper was to present the proximity of clothing to self(PCS) which is a subjective measure developed by Sontag and Lee(2004). PCS is the scale to examine the relationship between self and clothing. Since PCS scale was prepared for American consumers, it is necessary to verify the validity as a measure to Korean consumers. Consumers' quality of life are functionally related to satisfaction with all of life's domain and subdomains. Clothing in life facilitates the creativity and expressiveness to the wearer, especially to adolescent. The concept of bottom-up spillover effect developed by Lee et al.(2002) was adopted as research model. Data were gathered by surveying university students using convenient sampling, 419 questionnaires were used in the statistical analysis. It was proved that PCS scale was effective measure to Korean consumers. And also PCS scale could be a proxy indicator in view of the fact that clothing was deeply related to consumers' quality of life.