• Journal of Applied and Pure Mathematics
• /
• v.5 no.1_2
• /
• pp.69-93
• /
• 2023

Here we give multivariate quantitative approximations of Banach space valued continuous multivariate functions on a box or ℝ^N, N ∈ ℕ, by the multivariate normalized, quasi-interpolation, Kantorovich type and quadrature type neural network operators. We treat also the case of approximation by iterated operators of the last four types. These approximations are derived by establishing multidimensional Jackson type inequalities involving the multivariate modulus of continuity of the engaged function or its high order Fréchet derivatives. Our multivariate operators are defined by using a multidimensional density function induced by a parametrized Gudermannian sigmoid function. The approximations are pointwise and uniform. The related feed-forward neural network is with one hidden layer.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2020.11a
• /
• pp.949-952
• /
• 2020

층간 소음으로 인한 갈등, 직장에서 발생하는 소음으로 인한 청력 감소 등 소음으로 인한 많은 문제들이 최근 들어 많이 대두되고 있다. 소음을 없애기 위해서는 물리적인 방법도 있지만 ANC 기술을 이용하여 소음을 없애는 방법도 있다. ANC 란 기존 소음과 반대 위상인 파장을 출력해 상쇄간섭을 일으켜 소음을 줄여주는 기술이다. AI_ANC 기술은 기존에 ANC 기술의 한계점(Feedback, Time Gap)을 극복하기 위해 Feed forward 방식과 AI 를 접목시켜 능동적으로 소음을 감소시킨다. 이기술은 층간 소음, 공장 소음 더 나아가 다양한 이동수단까지 범용적으로 활용성이 뛰어나다.

• Annual Conference on Human and Language Technology
• /
• 2019.10a
• /
• pp.34-36
• /
• 2019

기존 딥 러닝을 이용한 형태소 분석 및 품사 태깅(Part-Of-Speech tagging)은 feed-forward neural network에 CRF를 결합하는 방법이나 sequence-to-sequence 모델을 이용한 방법 등의 다양한 모델들이 연구되었다. 본 논문에서는 한국어 형태소 분석 및 품사 태깅을 수행하기 위하여 최근 자연어처리 태스크에서 많은 성능 향상을 보이고 있는 BERT를 기반으로 한 음절 단위 LSTM-CRF 모델을 제안한다. BERT는 양방향성을 가진 트랜스포머(transformer) 인코더를 기반으로 언어 모델을 사전 학습한 것이며, 본 논문에서는 한국어 대용량 코퍼스를 어절 단위로 사전 학습한 KorBERT를 사용한다. 실험 결과, 본 논문에서 제안한 모델이 기존 한국어 형태소 분석 및 품사 태깅 연구들 보다 좋은 (세종 코퍼스) F1 98.74%의 성능을 보였다.

• Proceedings of the KIPE Conference
• /
• 2020.08a
• /
• pp.261-262
• /
• 2020

본 논문은 DC/DC 위상천이 풀브릿지 컨버터의 과도상태 개선을 위한 듀얼모드 제어를 제안한다. 기존의 discontinuous conduction mode (DCM)과 continuous conduction mode (CCM)을 이용하여 운용하는 위상천이 풀브릿지 컨버터는 각 모드의 상이한 전압게인의 특성을 고려하지 않아 급격한 로드 변동시 출력 전압의 언더슈트 혹은 오버슈트가 크게 발생한다. 제안하는 듀얼모드 제어는 DCM과 CCM 각 모드 별개로 제어가 적용되고 출력 전류의 변화에 대응하는 적응형 feed-forward 제어가 적용되어 컨버터의 과도상태를 개선할 수 있다. 듀얼모드의 제어 방법의 성능과 타당성을 시뮬레이션을 통해 입증한다.

• Journal of Advanced Information Technology and Convergence
• /
• v.9 no.2
• /
• pp.75-87
• /
• 2019

This paper presents a fully digital controlled power supply unit for the wireless controlled dimmable LED lighting system. The proposed power supply designed using a fly-back converter, which is directly controlled by a microprocessor. Although the proposed circuit does not sense the AC input current and has not AC input voltage feed-forward, it can achieve a high power factor. The proposed power supply directly regulates the output power for LED loads using the PWM and PFM control of the fly-back converter without additional regulator. For a wireless remote control function, the Zigbee modem is equipped in the proposed power supply. A prototype set-up has been built and tested. Through the experiment with a prototype set-up, the usefulness of the proposed power supply is verified.

• The Bulletin of The Korean Astronomical Society
• /
• v.46 no.2
• /
• pp.45.3-46
• /
• 2021

Supernovae classes have been defined phenomenologically, based on spectral features and time series data, since the specific details of the physics of the different explosions remain unrevealed. However, the number of these classes is increasing as objects with new features are observed, and the next generation of large-surveys will only bring more variety to our attention. We apply the machine learning technique of multi-label classification to the spectra of supernovae. By measuring the probabilities of specific features or 'tags' in the supernova spectra, we can compress the information from a specific object down to that suitable for a human or database scan, without the need to directly assign to a reductive 'class'. We use logistic regression to assign tag probabilities, and then a feed-forward neural network to filter the objects into the standard set of classes, based solely on the tag probabilities. We present STag, a software package that can compute these tag probabilities and make spectral classifications.

• Membrane Journal
• /
• v.28 no.5
• /
• pp.368-377
• /
• 2018

The n-nitrilotris(methylene) phosphonic acid (NTPA) potassium salt was synthesized as a draw solute for forward osmosis. NTPA-4K, NTPA-5K and NTPA-6K were synthesized by varying the content of KOH added to NTPA and confirmed by $^1H$-NMR and $^{13}C$-NMR. The osmotic pressure, viscosity, water flux and reverse salt flux were measured to characterize the draw solute. In the forward osmosis process when distilled water was used as a feed solution and 0.5 M of NTPA-4K, NTPA-5K and NTPA-6K were used as a draw solution, the water flux was 35.8, 38.8 and 42.2 LMH, the reverse salt flux was 5.4, 6.9 and 7.4 gMH, respectively. It was confirmed that the water flux was lower than the conventional NaCl draw solution, but the reverse salt flux was much lower. In order to recover the diluted draw solution, nanofiltration was conducted. The results showed that the draw solute could be retained by above 90%.

• Membrane Journal
• /
• v.32 no.4
• /
• pp.235-252
• /
• 2022

Forward osmotic pressure-free reverse osmosis (Δ𝜋=0 RO) was invented in 2013. The first patent (US 9,950,297 B2) was registered on April 18, 2018. The "Osmotic Pressure of Concentrated Solutions" in JACS (1908) by G.N. Lewis of MIT was used for the estimation. The Chang's RO system differs from conventional RO (C-RO) in that two-chamber system of osmotic pressure equalizer and a low-pressure RO system while C-RO is based on a single chamber. Chang claimed that all aqueous solutions, including salt water, regardless of its osmotic pressure can be separated into water and salt. The second patent (US 10.953.367B2, March 23, 2021) showed that a low-pressure reverse osmosis is possible for 3.0% input at Δ𝜋 of 10 to 12 bar. Singularity ZERO reverse osmosis from his third patent (Korea patent 10-22322755, US-PCT/KR202003595) for a 3.0% NaCl input, 50% more water recovery, use of 1/3 RO membrane area, and 1/5th of theoretical energy. These numbers come from Chang's laboratory experiments and theoretical analysis. Relative residence time (RRT) of feed and OE chambers makes Δ𝜋 to zero or negative by recycling enriched feed flow. The construction cost by S-ZERO was estimated to be around 50~60% of the current RO system.

• Korean Journal of Psychosomatic Medicine
• /
• v.9 no.1
• /
• pp.81-92
• /
• 2001

Stress has been linked to the pathophysiology and pathogenesis of various psychiatric illnesses. Over the past few years, our understanding of the brain and neuroendocrine systems that are linked to stress responses has increased enormously. This article reviews a series of animal and human studies to understand what are the central pathways by which stress is perceived, processed, and transduced into a neuroendocrine response. We focus on the limbic-hypothalamic-pituitary-adrenal(LHPA) axis and several neurotransmitter systems such as norepinephrine, CRF, serotonin, acetylcholine, and dopamine. LHPA stress circuit is a complex system with multiple control mechanisms which are altered in pathological states. CRF and related peptides in the central nervous system appear to enhance behavioral responses to stressors. Norepinephrine systems are also activated by stressors and cause the release of catecholamines from the autonomic nervous system. CRF-norepinephrine interaction makes a feed-forward system which may be important for an organism to mobilize not only the pituitary system but also the central nervous system, in response to environmental challenges. The interactions among several neurotransmitters and endocrine systems appear to play key roles in mediating various behavioral and psychological stress responses involving abnormal responses to stressors such as anxiety and affective disorders.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.42 no.2 s.332
• /
• pp.23-30
• /
• 2005

This paper describes a DLL(delay locked loop)-based multi-clock generator having the lower active stand-by power as well as a fast relocking after re-activating the DLL. for low power and high speed VLSI chip. It enables a frequency multiplication using frequency multiplier scheme and produces output clocks with 50:50 duty-ratio regardless of the duty-ratio of system clock. Also, digital control scheme using DAC enables a fast relocking operation after exiting a standby-mode of the clock system which was obtained by storing analog locking information as digital codes in a register block. Also, for a clock multiplication, it has a feed-forward duty correction scheme using multiphase and phase mixing corrects a duty-error of system clock without requiring additional time. In this paper, the proposed DLL-based multi-clock generator can provides a synchronous clock to an external clock for I/O data communications and multiple clocks of slow and high speed operations for various IPs. The proposed DLL-based multi-clock generator was designed by the area of $1796{\mu}m\times654{\mu}m$ using $0.35-{\mu}m$ CMOS process and has $75MHz\~550MHz$ lock-range and maximum multiplication frequency of 800 MHz below 20psec static skew at 2.3v supply voltage.