http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Magneto-transport Properties of GaMnAs:Si Ferromagnetic Semiconductors
Hyungchan Kim,Hakjoon Lee,정선재,Y. J. Cho,X. Liu,J. K. Furdyna,이상훈 한국물리학회 2009 THE JOURNAL OF THE KOREAN PHYSICAL SOCIETY Vol.55 No.1
The magnetic properties of a series of GaMnAs:Si ferromagnetic semiconductor films, in which the Mn concentration ranges from 7% to 10%, were investigated by using magneto-transport measurements. The temperature dependence of the resistivity revealed a systematic increase in the Curie temperature (Tc) with increasing Mn concentration in the series. Since the Tc of the undoped GaMnAs ferromagnetic semiconductor decreases with increasing Mn concentration above 6%, the observation of a systematic increase of Tc with increasing Mn concentration in our GaMnAs:Si series indicates the effectiveness of our counter doping for the incorporation of a a large amount of 7% Mn in the system. The field scan of the planar Hall effect (PHE) showed a typical two-step switching behavior at low temperatures, indicating the presence of a strong cubic anisotropy. The switching fields, however, systematically decreased with increasing Mn concentration in the series. The angular dependences of the switching fields were fitted by using the magnetic free energy and Cowburn’s model to obtained the domain pinning energy, which showed systematically smaller values as the Mn concentration of the sample was increased. The temperature dependences of the pinning energies indicated a change in the uniaxial anisotropy from the [110] to the [110] direction with increasing Mn concentration in the series.
심층 신경망 기반 단채널 음성 향상을 위한 SNR 에 따른 손실 함수 강조 기법
강경묵(Kyeongmuk Kang),송형찬(Hyungchan Song),이은균(Eunkyun Lee),신종원(Jong Won Shin) 한국통신학회 2021 한국통신학회 학술대회논문집 Vol.2021 No.11
본 논문에서는 심층 신경망 기반의 단채널 음성 향상에서 음성의 신호 대 잡음 비(Signal to Noise ratio, 이하 SNR)에 따른 손실 함수 강조 기법을 제안한다. 모든 SNR 에 대해 동일한 손실 함수를 적용하는 기존의 방법과 달리 심층신경망의 손실 함수에 제안한 변조 항을 적용해주어 SNR 에 따른 차등 강조를 적용하였다. 상대적으로 음성을 추정하기 어려운 샘플에 손실 함수를 더 강조해주어 낮은 SNR 일 때 기존 방법보다 더욱 효과적으로 음성 신호의 품질과 명료도를 증진시킬 수 있다. 본 논문에서는 제안된 손실 함수 강조 기법을 적용한 심층 신경망을 이용해 얻어진 음성과 기존의 손실 함수만 사용했을 때 얻어진 음성을 PESQ (perceptual evaluation of speech quality)를 이용하여 비교하였다.
RNN-LSTM 기반 공휴일 정보를 고려한 단기 전력수요예측
김한솔(Hansol Kim),송형찬(Hyungchan Song),고석갑(Seok-Kap Ko),이병탁(Byung-Tak Lee),신종원(Jong Won Shin) 대한전자공학회 2016 대한전자공학회 학술대회 Vol.2016 No.11
Daily electricity demand and its fluctuation have increased by abrupt climate change and excessive use of air conditioning and these has affected to forecast the short-term electricity load. Also, the electricity load pattern learning is disturbed by holidays that cause sudden the electricity demand reduction. We proposed the feature extraction algorithm for demand reduction in holidays and implemented the RNN-LSTM (Recurrent Neural Network-Long Short Term Memory) based forecasting. The results were compared with the forecasting performance of SARIMA (Seasonal Auto Regressive Integrated Moving Average). The comparative result shows that RNN-LSTM outperforms SARIMA.