안성진 교수
Sungjin Ahn
성균관대학교 컴퓨터교육과 · 컴퓨터과학
연구실 소개
안성진 교수의 연구실은 그래프 신경망과 그래프 자동에코더 기반의 비지도 학습 기법을 활용해 복잡한 네트워크 구조에서의 링크 예측 문제를 해결하는 데 초점을 맞추고 있습니다. 특히 고립 노드가 포함된 상황에서도 정확한 임베딩을 도출할 수 있도록 L2 정규화를 통합한 새로운 변동형 그래프 자동에코더(VGNAE)를 개발하며, 그래프 기반 임베딩의 정밀도를 향상시키는 데 기여하고 있습니다. 또한, 2D 반도체 시스템에서의 전자 운동성과 열역학적 거동을 이론적으로 분석함으로써, 고이동도 시스템에서의 유체역학적 행동과 온도-전도도 비율 위반 현상 등을 규명하고 있습니다. 이와 더불어 고해상도 디스플레이의 저전력 스캔 드라이버 설계 및 신호 분류 알고리즘 개선 등 실용적 응용 기술 개발에도 기여하고 있습니다.
연구 현황
연구 성과 추이
표시된 성과는 수집된 데이터 기준으로 산출되며, 일부 차이가 있을 수 있습니다.
주요 논문
15Link prediction is one of the key problems for graph-structured data. With the advancement of graph neural networks, graph autoencoders (GAEs) and variational graph autoencoders (VGAEs) have been proposed to learn graph embeddings in an unsupervised way. It has been shown that these methods are effective for link prediction tasks. However, they do not work well in link predictions when a node whose degree is zero (i.g., isolated node) is involved. We have found that GAEs/VGAEs make embeddings of
Considering theoretically the transition between hydrodynamic and ballistic regimes in 2D semiconductors, we show that electrons in high-mobility 2D GaAs are by far the best system for the direct observation of collective hydrodynamic effects even in bulk transport properties independent of complicated transport features in narrow constrictions and small systems where Gurzhi phenomena are typically studied experimentally. We predict a strong hydrodynamics-induced generic violation of the Wiedema
Electrons (or holes) confined in two-dimensional (2D) semiconductor layers have served as model systems for studying disorder and interaction effects for almost 50 years. In particular, strong disorder drives the metallic 2D carriers into a strongly localized Anderson insulator (AI) at low densities whereas pristine 2D electrons in the presence of no (or little) disorder should solidify into a Wigner crystal (WC) at very low carrier densities in order to optimize their Coulomb potential energy.
This paper proposes M-ary circular θ-quadrature amplitude modulation (CTQAM) for M = 2 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">l</sup> , l ≥ 4, which is constructed by rearrangement of the signal points based on θ-QAM. We provide a construction method for the signal constellation of CTQAM and analyze the symbol and bit error performances of M-ary CTQAM in additive white Gaussian noise (AWGN) and Nakagami-m fading channels. Through computer s
Abstract In this paper, a low‐power scan driver using depletion‐mode amorphous indium‐gallium‐zinc‐oxide (a‐IGZO) thin‐film transistors (TFTs) is proposed for high‐resolution displays. The proposed scan driver uses single clock signal to reduce the power consumption. This scan driver is simulated at the operating frequency of 153.6 kHz, which is the driving condition of 10.0‐inch wide quadruple extended graphics array (WQXGA, 1600 × 2560) display with the frame frequency of 60 Hz. The power cons
We propose an enhanced automatic modulation classification algorithm based on Kolmogorov-Smirnov test. The proposed classifier employs the real and imaginary components extracted from the received signal as separate decision statistics. Also, unlike the conventional K-S test based algorithm, mean square error (MSE) between the empirical cumulative distribution and the hypothesized distribution for each modulation candidate is evaluated in the proposed algorithm. Simulation results show that the
Hierarchical modulation provides unequal error protection to transmitted data according to their relative importance. In this article, hierarchical modulation constellation based on circular θ-quadrature amplitude modulation (CTQAM), 4/M-CTQAM, M = 2 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">k</sup> , k ≥ 4, is proposed. We provide a construction method and bit-to-symbol mapping for the proposed hierarchical modulation constellation. We then e
This article is aimed at defining items of analysis using SNMP MIB-II for the purpose of analyzing the performance of Internet-based networks running on TCP/IP protocol, and then utilizing these items, in conjunction with various Web technology and JAVA, to design and implement a Web-based interface of a management system to analyze the performance of the Internet. Copyright © 1999 John Wiley & Sons, Ltd.
Monolayer (ML) molybdenum disulfide (MoS2) is a promising material for next-generation optoelectronic applications because MoS2 exhibits remarkable electronic and optical properties owing to its flexibility and direct bandgap. Despite these outstanding properties, MoS2 has limitations in fabricating optoelectronic devices because of its low quantum yield. To overcome these limitations, we propose a photoluminescence (PL) enhancement technique using a surface plasmon effect induced by transferrin
M-ary stepped θ-QAM for M = 2 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2l</sup> , m ≥ 3, was reported recently. Symbol error rate (SER) and optimum angle θ in terms of minimum SER were analyzed and it was confirmed that optimum angle converges to 60°. However bit error rate (BER) of stepped θ-QAM has not been analyzed yet. In this paper, we present a bit-to-symbol mapping by using layer labeling algorithm which is computationally efficient me
Introduction This study investigates support strategies to enhance basic software education as a liberal arts course at Konkuk University, South Korea, by integrating design thinking and computational thinking to mitigate the cognitive load of programming. Methods A study with 190 students utilized a learner-centered approach, incorporating design thinking and computational problem-solving, and evaluated its effectiveness through pre-and post-tests and structural equation modeling. Results The i
대표 연구 분야
안성진 교수의 연구를 Nubint에서 더 깊이 살펴보세요
이 연구실의 논문을 앱에서 열어 AI와 함께 읽고, 핵심을 요약하고, 내 글에 인용하세요.