• Journal of Information Technology Applications and Management
• /
• 제29권3호
• /
• pp.43-55
• /
• 2022

In this study, with the goal of developing a deep learning-based product recommendation model for effective matching of influencers and products, a deep learning model with a collaborative filtering model combined with generalized matrix decomposition(GMF), a collaborative filtering model based on multi-layer perceptron (MLP), and neural collaborative filtering and generalized matrix Factorization (NeuMF), a hybrid model combining GMP and MLP was developed and tested. In particular, we utilize one-class problem free boosting (OCF-B) method to solve the one-class problem that occurs when training is performed only on positive cases using implicit feedback in the deep learning-based collaborative filtering recommendation model. In relation to model selection based on overall experimental results, the MLP model showed highest performance with weighted average precision, weighted average recall, and f1 score were 0.85 in the model (n=3,000, term=15). This study is meaningful in practice as it attempted to commercialize a deep learning-based recommendation system where influencer's promotion data is being accumulated, pactical personalized recommendation service is not yet commercially applied yet.

• Korean Journal of Radiology
• /
• 제23권5호
• /
• pp.539-547
• /
• 2022

Objective: To investigate the association between functional tumor burden of peritoneal carcinomatosis (PC) derived from diffusion-weighted imaging (DWI) and overall survival in patients with advanced ovarian carcinoma (OC). Materials and Methods: This prospective study was approved by the local research ethics committee, and informed consent was obtained. Fifty patients (mean age ± standard deviation, 57 ± 12 years) with stage III-IV OC scheduled for primary or interval debulking surgery (IDS) were recruited between June 2016 and December 2021. DWI (b values: 0, 400, and 800 s/mm²) was acquired with a 16-channel phased-array torso coil. The functional PC burden on DWI was derived based on K-means clustering to discard fat, air, and normal tissue. A score similar to the surgical peritoneal cancer index was assigned to each abdominopelvic region, with additional scores assigned to the involvement of critical sites, denoted as the functional peritoneal cancer index (fPCI). The apparent diffusion coefficient (ADC) of the largest lesion was calculated. Patients were dichotomized by immediate surgical outcome into high- and low-risk groups (with and without residual disease, respectively) with subsequent survival analysis using the Kaplan-Meier curve and log-rank test. Multivariable Cox proportional hazards regression was used to evaluate the association between DWI-derived results and overall survival. Results: Fifteen (30.0%) patients underwent primary debulking surgery, and 35 (70.0%) patients received neoadjuvant chemotherapy followed by IDS. Complete tumor debulking was achieved in 32 patients. Patients with residual disease after debulking surgery had reduced overall survival (p = 0.043). The fPCI/ADC was negatively associated with overall survival when accounted for clinicopathological information with a hazard ratio of 1.254 for high fPCI/ADC (95% confidence interval, 1.007-1.560; p = 0.043). Conclusion: A high DWI-derived functional tumor burden was associated with decreased overall survival in patients with advanced OC.

• 전자공학회논문지T
• /
• 제36T권4호
• /
• pp.17-23
• /
• 1999

본 연구는 문맥 독립형 화자 인식 알고리즘을 이용하여 보안시스템을 소프트웨어와 하드웨어로 구성한 논문이다. 화자인식을 이용한 보안시스템은 윈도우상에서 사운드카드를 이용하여 음성을 입력받고, 성도 모델링을 이용한 음성 파라미터를 추출하였으며, k-means 클러스터링 알고리즘을 기반으로 하여 화자를 모델링하였다. 등록된 화자에 대한 인식된 결과는 PIC16F84 마이크로 프로세서를 이용하여 자물쇠를 개${\cdot}$폐하도록 구성하였다. OFF-LINE의 실험은 TIMIT데이터를 이용하였으며, 5명의 화자에 대하여 ON-LINE으로 인식한 결과 학습시킨 데이터에 대해서는 100%의 인식률을 얻었으며 학습을 시키지 않은 데이터에 대해서는 99%의 인식률을 얻었다. 그리고 사용자 거부율 1%, 사칭자 허용률 0%, 검증평균오류는 0.5%를 보였다.

• Safety and Health at Work
• /
• 제8권1호
• /
• pp.105-115
• /
• 2017

Background: The goal of this study is to develop a general population job-exposure matrix (GPJEM) on asbestos to estimate occupational asbestos exposure levels in the Republic of Korea. Methods: Three Korean domestic quantitative exposure datasets collected from 1984 to 2008 were used to build the GPJEM. Exposure groups in collected data were reclassified based on the current Korean Standard Industrial Classification ($9^{th}$ edition) and the Korean Standard Classification of Occupations code ($6^{th}$ edition) that is in accordance to international standards. All of the exposure levels were expressed by weighted arithmetic mean (WAM) and minimum and maximum concentrations. Results: Based on the established GPJEM, the 112 exposure groups could be reclassified into 86 industries and 74 occupations. In the 1980s, the highest exposure levels were estimated in "knitting and weaving machine operators" with a WAM concentration of 7.48 fibers/mL (f/mL); in the 1990s, "plastic products production machine operators" with 5.12 f/mL, and in the 2000s "detergents production machine operators" handling talc containing asbestos with 2.45 f/mL. Of the 112 exposure groups, 44 groups had higher WAM concentrations than the Korean occupational exposure limit of 0.1 f/mL. Conclusion: The newly constructed GPJEM which is generated from actual domestic quantitative exposure data could be useful in evaluating historical exposure levels to asbestos and could contribute to improved prediction of asbestos-related diseases among Koreans.

• 음성과학
• /
• 제5권1호
• /
• pp.7-21
• /
• 1999

This paper proposes an improved training procedure in speech recognition based on the continuous density of the Hidden Markov Model (CDHMM). Of the three parameters (initial state distribution probability, state transition probability, output probability density function (p.d.f.) of state) governing the CDHMM model, we focus on the third parameter and propose an efficient algorithm that determines the p.d.f. of each state. It is known that the resulting CDHMM model converges to a local maximum point of parameter estimation via the iterative Expectation Maximization procedure. Specifically, we propose two independent algorithms that can be embedded in the segmental K -means training procedure by replacing relevant key steps; the adaptation of the number of mixture Gaussian p.d.f. and the initialization using the CDHMM parameters previously estimated. The proposed adaptation algorithm searches for the optimal number of mixture Gaussian humps to ensure that the p.d.f. is consistently re-estimated, enabling the model to converge toward the global maximum point. By applying an appropriate threshold value, which measures the amount of collective changes of weighted variances, the optimized number of mixture Gaussian branch is determined. The initialization algorithm essentially exploits the CDHMM parameters previously estimated and uses them as the basis for the current initial segmentation subroutine. It captures the trend of previous training history whereas the uniform segmentation decimates it. The recognition performance of the proposed adaptation procedures along with the suggested initialization is verified to be always better than that of existing training procedure using fixed number of mixture Gaussian p.d.f.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송공학회 2002년도 정기총회 및 학술대회
• /
• pp.3-8
• /
• 2002

A new Modem technique - 8-state Superposed QAM(8SQAM) - for use in power and bandwidth limited digital communication system is proposed. The premodulation baseband signal of the 8SQAM system is a double-interval(2Ts) raised cosine pulse superposed with weighted single-interval(Ts) raised cosine pulse. The performance of the 8SQAM, in a nonlinearly amplified multi-channel environment with additive white Gaussian noise(AWGN), inter-symbol interference(ISI), inter-modulation(IM) and adjacent channel interference(ACI), is analyzed via computer simulation. The simulation result shows that the 8SQAM outperforms 8PSK by about 1.6㏈ at BER=$1\times10^{-4}$ when HPA IBO is 1㏈ and channel Spacing, ${\Delta}F$ is 0.417fb.

• 호남수학학술지
• /
• 제41권1호
• /
• pp.51-65
• /
• 2019

In this paper, we give sufficient conditions to guarantee the asymptotic stability of the zero solution to a kind of nonlinear fractional dynamic equations of order ${\alpha}$ (1 < ${\alpha}$ < 2). By using the Krasnoselskii's fixed point theorem in a weighted Banach space, we establish new results on the asymptotic stability of the zero solution provided f (t, 0) = 0, which include and improve some related results in the literature.

• 한국방사선학회논문지
• /
• 제13권1호
• /
• pp.111-118
• /
• 2019

흉부 디지털 단층영상합성장치는 기존 DR의 낮은 깊이 해상도, CT의 높은 피폭선량 문제를 해결할 수 있는 획기적인 영상장치로 대두되고 있다. 그러나 제한된 스캔 각도로 인해 프로젝션이 X 선 소스 동작 방향으로 흉부를 완전히 포함 할 수 없어 재구성 된 슬라이스의 위, 아래 방향 경계를 따라 강도의 불연속성이 발생하게 되는데 이러한 현상을 잘림 아티팩트 (Truncation artifact)라고 한다. 이 연구의 목적은 가중 정규화 접근법을 사용하여 잘림 아티팩트를 줄이고 리스템에서 개발한 프로토 타입 흉부 디지털 단층영상합성장치 시스템에 대한 이 접근법의 성능을 평가하는 것이다. 이 시스템의 source-to-image distance는 1100 mm 이고 X 선원의 회전 중심은 검출기 표면에서 100mm 위로 설정되었다. LUNGMAN 팬텀을 사용하여 ${\pm}20^{\circ}$의 투영 뷰를 $1^{\circ}$ 간격으로 41장을 얻은 후, filtered back projection 알고리즘으로 재구성했다. 정량적 평가를 위하여 시뮬레이션을 이용하여 기준영상을 재구성 후 peak signal to noise ratio와 structure similarity index 값을 평가하였으며 실제 실험 데이터를 이용하여 mean value of specific direction 값을 평가하였다. 시뮬레이션 결과로 아티팩트 보정 전 일반적인 filtered back projection 알고리즘으로 재구성 한 영상과 비교하여 peak signal to noise ratio값과 structure similarity index값 모두 각각 증가하였으며, 실제 실험 재구성 영상의 mean value of specific direction 결과는 아티팩트의 영향이 감소됨을 확인할 수 있었다. 결론적으로, 가중 정규화 방법은 잘림 아티팩트를 줄임으로써 진단의 어려움을 발생시키는 가능성을 개선시킬 수 있는 방법으로 사료된다.

• Industrial Engineering and Management Systems
• /
• 제13권1호
• /
• pp.15-28
• /
• 2014

Evaluation of the implementation and performance of a firm's green supply chain management (GSCM) is an ongoing process. Balanced scorecard is a multi-criteria evaluation concept that highlights implementation and performance measures. The literature on the framework is abundant literature but scarce on how to build a hierarchical framework under uncertainty with dependence relations. Hence, this study proposes a hybrid approach, which includes applied interpretive structural modeling to build a hierarchical structure and uses the analytic network process to analyze the dependence relations. Additionally, this study applies the fuzzy set theory to determine linguistic preferences. Twenty dependence criteria are evaluated for a GSCM implemented firm in Taiwan. The result shows that the financial aspect and life cycle assessment are the most important performance and weighted criteria.

• Investigative Magnetic Resonance Imaging
• /
• 제1권1호
• /
• pp.79-85
• /
• 1997

Fast spin echo imaging utilizes multiple spin echoes to encode multiple k-space lines instead of multiple $T_2-weighted$ images. As results, intensities in k-space data are varying according to T2 decay, which generates Gibb's artifact in the reconstructed image. The echo time for e encoding dc block determines contrast, as is specified by the effective echo time, however, all location of other echoes to different k-space frequency blocks in fast spin echo imaging is not f fully investigated. In this study, symmetric arrangement of multiple echoes in k-space is investigated to reduce Gibb's artifact. Design of filters based on the measurement of multiple e echo intensities is also proposed in two stage manner, i.e., equalization and filtering. From s simulation and experiment, it was observed that Gibb's phenomena were substantially reduced b by the proposed methods.