• Journal of Information Processing Systems
• /
• 제15권1호
• /
• pp.91-99
• /
• 2019

An innovative tone modeling framework based on deep neural networks in tone recognition was proposed in this paper. In the framework, both the prosodic features and the articulatory features were firstly extracted as the raw input data. Then, a 5-layer-deep deep belief network was presented to obtain high-level tone features. Finally, support vector machine was trained to recognize tones. The 863-data corpus had been applied in experiments, and the results show that the proposed method helped improve the recognition accuracy significantly for all tone patterns. Meanwhile, the average tone recognition rate reached 83.03%, which is 8.61% higher than that of the original method.

• International Journal of Advanced Culture Technology
• /
• 제8권4호
• /
• pp.119-128
• /
• 2020

Object detection is an important field of computer vision and is applied to applications such as security, autonomous driving, and face recognition. Recently, as the application of artificial intelligence technology including deep learning has been applied in various fields, it has become a more powerful tool that can learn meaningful high-level, deeper features, solving difficult problems that have not been solved. Therefore, deep learning techniques are also being studied in the field of object detection, and algorithms with excellent performance are being introduced. In this paper, a deep learning-based object detection algorithm used to detect multiple objects in an image is investigated, and future development directions are presented.

• Journal of Information Processing Systems
• /
• 제20권4호
• /
• pp.458-476
• /
• 2024

To address the issue of class imbalance in network traffic data, which affects the network intrusion detection performance, a combined framework using transformers is proposed. First, Tomek Links, SMOTE, and WGAN are used to preprocess the data to solve the class-imbalance problem. Second, the transformer is used to encode traffic data to extract the correlation between network traffic. Finally, a hybrid deep learning network model combining a bidirectional gated current unit and deep neural network is proposed, which is used to extract long-dependence features. A DNN is used to extract deep level features, and softmax is used to complete classification. Experiments were conducted on the NSLKDD, UNSWNB15, and CICIDS2017 datasets, and the detection accuracy rates of the proposed model were 99.72%, 84.86%, and 99.89% on three datasets, respectively. Compared with other relatively new deep-learning network models, it effectively improved the intrusion detection performance, thereby improving the communication security of network data.

• 정보과학회 컴퓨팅의 실제 논문지
• /
• 제23권5호
• /
• pp.299-303
• /
• 2017

딥 러닝(Deep learning)은 기존 인공 신경망 내 계층 수를 증가시킴과 동시에 효과적인 학습 방법론을 제시함으로써 객체/음성 인식 및 자연어 처리 등 고수준 문제 해결에 있어 괄목할만한 성과를 보이고 있다. 그러나 학습에 필요한 시간과 리소스가 크다는 한계를 지니고 있어, 이를 줄이기 위한 연구가 활발히 진행되고 있다. 본 연구에서는 아파치 스파크 기반 클러스터 컴퓨팅 프레임워크 상에서 딥 러닝을 분산화하는 두 가지 툴(DeepSpark, SparkNet)의 성능을 학습 정확도와 속도 측면에서 측정하고 분석하였다. CIFAR-10/CIFAR-100 데이터를 사용한 실험에서 SparkNet은 학습 과정의 정확도 변동 폭이 적은 반면 DeepSpark는 학습 초기 정확도는 변동 폭이 크지만 점차 변동 폭이 줄어들면서 SparkNet 대비 약 15% 높은 정확도를 보였고, 조건에 따라 단일 머신보다도 높은 정확도로 보다 빠르게 수렴하는 양상을 확인할 수 있었다.

• 한국지하수토양환경학회지:지하수토양환경
• /
• 제7권3호
• /
• pp.71-77
• /
• 2002

본 연구는 비교적 인위적 오염의 개연성이 있다고 판단되는 난지도 주변 지역 토양을 대상으로 2001년 4월부터 5월까지 표토에서 깊이 3m까지 표토, 표토 1m깊이, 표토 2m깊이, 표토 3m깊이로 하여 중금속 오염농도를 조사 분석하였다. 조사대상 중금속은 Cd, Cu, Pb, As, $Cr^{+6}$ Hg의 6개 항목이였으며 농도 정량은 원자흡광광도 방법을 사용하였다. 결과로는 난지도 주변지역에서 중금속 평균도가 Cd 0.229mg/kg. Cu는 8.349mg/kg, Pb 11.083mg/kg, As 0.298mg/kg, $Cr^{+6}$ 0.124mg/kg, Hg 0.134mg/kg로 나타났다. 표토 평균농도는 Cd 0.305mg/kg, Cu 8.464mg/kg, Pb 11.383mg/kg, As 0.128mg/kg, $Cr^{+}$60.153mg/kg, and Hg 0.092mg/kg로 나타났다. 이는 Cd 과 $Cr^{+6}$가 조사대상 전체 평균농도의 80%수준으로 나타났다. Hg과 As는 전체 평균농도와 비슷한 수준으로 나타났다. 조사대상을 중심으로 살펴본 난지도 주변 지역의 중금속 농도조사에서 깊이별 변화 추이의 일관성은 찾을 수 없었으나 C지점과 D지점에서의 금속 수준은 오염된 향동천의 영향과 매립된 폐기물에 영향을 받은 것으로 생각되었다.

• 한국전기전자재료학회:학술대회논문집
• /
• 한국전기전자재료학회 2001년도 기술교육위원회 창립총회 및 학술대회 의료기기전시회
• /
• pp.52-59
• /
• 2001

The photoluminescence and the photo-current from p-type GaN films were investigated on both room- and low-temperatures for various Mg doping concentrations. At a low Mg doping level, there exists a photoluminescence center of the donor and the acceptor pair transition of the 3.28-eV band. This center is correlated with the defects for a shallow donor of the VGa and for an acceptor of MgGa. The acceptor level shows the binding energy of 0.2-0.25 eV, which was observed by the photon energy of the photo-current signal of 3.02-3.31 eV. At a high Mg doping level, there is a photoluminescence center of a deep donor and an acceptor pair transition of the 2.76-eV blue band. This center is attributed to the defect structures of MgGa-VN for the deep donor and MgGa for the acceptor. For low. doped samples, thermal annealing provides an additional photo-current signal for an unoccupied deep acceptor levels of 0.87-1.35 eV above valence band, indicating the p-type activation.

• 전자공학회논문지D
• /
• 제36D권5호
• /
• pp.63-68
• /
• 1999

본 논문에서는 저압 유기 금속 화학증착법으로 성장시킨 GaN박막들을 실온 케소드루미네슨스 방법으로 광학적 특성을 측정하여 결정성장 메커니즘과 광학적 특성과의 관계를 규명하였다. 관측된 스펙트럼은 주로 364nm의 강한 band-edge emission 피크와 550nm의 깊은 준위 피크이었다. 빔 전류의 증가에 따라 364nm 스펙트럼의 세기가 깊은 준위 발광 스펙트럼보다 크게 증가시켰다. 이는 성장 초기 GaN박막의 결정 결함이 깊은 준위 발광 스펙트럼과 깊은 관계가 있음을 나타내 주고 있다. 또한 미세 결정 구조와 깊은 준위 발광 스펙프럼과의 관계 분석을 위해 주사형 전자현미경 사진과 캐소드루미네슨스 스펙트럼을 비교 검토하였다.

• The Journal of Asian Finance, Economics and Business
• /
• 제8권2호
• /
• pp.685-695
• /
• 2021

This study explores the impact of stochastic volatility in option pricing. To be more specific, we compare the option pricing performance between stochastic volatility option pricing model, namely, Heston option pricing model and standard Black-Scholes option pricing. Our finding, based on the market price of SET50 index option between May 2011 and September 2020, demonstrates stochastic volatility of underlying asset return for all level of moneyness. We find that both deep in the money and deep out of the money option exhibit higher volatility comparing with out of the money, at the money, and in the money option. Hence, our finding confirms the existence of volatility smile in Thai option markets. Further, based on calibration technique, the Heston option pricing model generates smaller pricing error for all level of moneyness and time to expiration than standard Black-Scholes option pricing model, though both Heston and Black-Scholes generate large pricing error for deep-in-the-money option and option that is far from expiration. Moreover, Heston option pricing model demonstrates a better pricing accuracy for call option than put option for all level and time to expiration. In sum, our finding supports the outperformance of the Heston option pricing model over standard Black-Scholes option pricing model.

• Nuclear Engineering and Technology
• /
• 제55권1호
• /
• pp.201-208
• /
• 2023

The dependence of the electrical characteristics on the fast neutron fluence of an epitaxial 4H-SiC Schottky barrier diode (SBD) was investigated. The 30 MeV cyclotron was used for fast neutron irradiation. The neutron fluences evaluated through Monte Carlo simulation were in the 2.7 × 10¹¹ to 1.45 × 10¹³ neutrons/cm² range. Current-voltage and capacitance-voltage measurements were performed to characterize the samples by extracting the parameters of the irradiated SBDs. Neutron-induced defects in the epitaxial layer were identified and quantified using a deep-level transient spectroscopy measurement system developed at the Korea Atomic Energy Research Institute. As the neutron fluence increased from 2.7 × 10¹¹ to 1.45 × 10¹³ neutrons/cm², the concentration of the Z_1/2 defects increased by approximately 20 times. The maximum defect concentration was estimated as 1.5 × 10¹⁴ cm^-3 at a neutron fluence of 1.45 × 10¹³ neutrons/cm².

• 산업경영시스템학회지
• /
• 제47권1호
• /
• pp.9-19
• /
• 2024

Deep learning-based computer vision anomaly detection algorithms are widely utilized in various fields. Especially in the manufacturing industry, the difficulty in collecting abnormal data compared to normal data, and the challenge of defining all potential abnormalities in advance, have led to an increasing demand for unsupervised learning methods that rely on normal data. In this study, we conducted a comparative analysis of deep learning-based unsupervised learning algorithms that define and detect abnormalities that can occur when transparent contact lenses are immersed in liquid solution. We validated and applied the unsupervised learning algorithms used in this study to the existing anomaly detection benchmark dataset, MvTecAD. The existing anomaly detection benchmark dataset primarily consists of solid objects, whereas in our study, we compared unsupervised learning-based algorithms in experiments judging the shape and presence of lenses submerged in liquid. Among the algorithms analyzed, EfficientAD showed an AUROC and F1-score of 0.97 in image-level tests. However, the F1-score decreased to 0.18 in pixel-level tests, making it challenging to determine the locations where abnormalities occurred. Despite this, EfficientAD demonstrated excellent performance in image-level tests classifying normal and abnormal instances, suggesting that with the collection and training of large-scale data in real industrial settings, it is expected to exhibit even better performance.