http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Implementing Action Mask in Proximal Policy Optimization (PPO) Algorithm
Cheng-Yen Tang,Chien-Hung Liu,Woei-Kae Chen,Shingchern D. You 한국통신학회 2020 ICT Express Vol.6 No.3
The proximal policy optimization (PPO) algorithm is a promising algorithm in reinforcement learning. In this paper, we propose to add an action mask in the PPO algorithm. The mask indicates whether an action is valid or invalid for each state. Simulation results show that, when compared with the original version, the proposed algorithm yields much higher return with a moderate number of training steps. Therefore, it is useful and valuable to incorporate such a mask if applicable.
Cheng, Chun-An,Chang, Chien-Hsuan,Cheng, Hung-Liang,Chung, Tsung-Yuan,Tseng, Ching-Hsien,Tseng, Kuo-Ching The Korean Institute of Power Electronics 2016 JOURNAL OF POWER ELECTRONICS Vol.16 No.4
This study proposes a single-stage light-emitting diode (LED) tube lamp driver with input-current shaping for T8/T10-type fluorescent lamp replacements. The proposed AC-DC LED driver integrates a dual-boost converter with coupled inductors and a half-bridge series-resonant converter with a bridge rectifier into a single-stage power conversion topology. This paper presents the operational principles and design considerations for one T8-type 18 W-rated LED tube lamp with line input voltages ranging from 100 V rms to 120 V rms. Experimental results for the prototype driver show that the highest power factor (PF = 0.988), lowest input current total harmonic distortion (THD = 7.22%), and highest circuit efficiency (η = 92.42%) are obtained at an input voltage of 120 V. Hence, the proposed driver is feasible for use in energy-efficient indoor lighting applications.
A Simple Partial Discharge Detector for Low-Voltage Rotating Electrical Machines
Cheng-Chi Tai,Ting-Cheng Huang,Ching-Chau Su,Chien-Yi Chen,Ju-Chu Hsieh,Yu-Shiun Lin,Chung-Tzong Wang,Jeng-Hung Lai 전력전자학회 2007 ICPE(ISPE)논문집 Vol.- No.-
A new, simple partial discharge (PD) detector for low-voltage rotating electrical machines using acoustical emission (AE) technique is dseveloped in this study. Common electric components were used in the detector, which reduces the cost of inspection, comparing with the traditional PD detection methods that use expensive equipment costing from tens of thousands to several millions dollars. Experimental results by resonant type AE sensors (150 ㎑) which utilize power line-cycle in a microcontroller unit (MCU) as reference to measure the PDs generated in a low-voltage motor are presented. The AE signals are then amplified by a pre-amplifier (30 ㎑ ~ 300 ㎑, 34 ㏈). Since the resonant frequency of the sensor is much lower than that of the electromagnetic (EM) interferences around the motor, the effects of noise is substantially reduced by this method. In the mean time, the use of 150-㎑ resonant type AE sensor also avoids the disturbance of mechanical vibration noise. According to the experiment results, the measurement system developed in this study can be used to detect the PDs’ AE signals correctly. The AE measurement scheme proposed in this study provides an effective, low-cost method for PD measurements.
On the Maximum Buffer Size Achieved in a Class of Constructions of Optical Priority Queues
Cheng, Jay,Huang, Shin-Shiang,Chou, Hsin-Hung,Tang, Ming-Che 한국통신학회 2023 Journal of communications and networks Vol.25 No.4
The design of optical buffers is an important issuein all-optical packet switching. One of the most general types ofbuffering schemes is priority queues, which includes first-in first-out (FIFO) queues and last-in first-out (LIFO) queues as specialcases (where the packet arrival times are used for the assignmentof packet priorities). Recently, it was shown in our previous workthat an optical priority queue with buffer size 2O(√αM) can beimplemented by using an optical (M + 2) × (M + 2) (bufferless)crossbar switch and M fiber delay lines under a simple priority-based routing policy, where α is a constant that depends onthe parameters used in the constructions. This achieved buffersize 2O( √αM) (which is exponential in √M) is the best resultcurrently known in the literature and significantly improveson all previous results (all of which are only polynomial inM). In this paper, we focus on our previous constructions ofoptical priority queues. The first contribution of this paper is toderive a closed-form expression for the maximum buffer sizethat can be achieved in our previous constructions. Such anexpression is of sufficient theoretical interest itself and can beused to directly compute the maximum buffer size (in contrast,the maximum buffer size has to be computed recursively inour previous work). The second contribution of this paper isto use the closed-form expression to show that in the regimethat s ≥ 2, k ≥ 2s + 1, and m ≥ 2, where s, k, and m areparameters used in the constructions, the maximum buffer sizeUk is given by Uk = 2O(√M log2 (2s+2) log2 m/((2s+1)m)) undera mild constraint that is applicable in practical scenarios. Thisresult can be regarded as a complement to the approximate resultUk ≈ 2O( √M log2 (2s+2) log2 m/((2s+1)m)) in our previous work.
Cheng Wei Lu,Chi-Feng Hung,Tzu Yu Lin,Ting Yang Hsieh,Su-Jane Wang 한국식품영양과학회 2019 Journal of medicinal food Vol.22 No.7
Evidence indicates that indirect inhibitory regulation of glutamatergic transmission, via reducing glutamate release, may induce neuroprotection. The present work was designed to examine whether allicin, a major component of garlic with neuroprotective effects, affected the release of glutamate evoked by 4-aminopyridine in rat cerebrocortical nerve terminals (synaptosomes). Allicin caused a potent inhibition on the release of glutamate evoked by 4-aminopyridine, and this inhibitory effect was abolished in the presence of Ca2+-free medium and vesicular transporter inhibitor. Allicin decreased the 4-aminopyridine-evoked elevation of intrasynaptosomal Ca2+ levels, but had no effect on the synaptosomal plasma membrane potential. The allicin-mediated inhibition of glutamate release was prevented by the N- and P/Q-type channel blocker and the protein kinase C (PKC) inhibitor, but was not affected by the intracellular Ca2+-release inhibitors, mitogen-activated protein kinase inhibitor, and protein kinase A inhibitor. Western blotting data also showed that allicin significantly reduced the phosphorylation of PKC. Together, these data indicate that in rat cerebrocortical nerve terminals, allicin depresses glutamate release and appears to decrease N- and P/Q-type Ca2+ channel and PKC activity.
Cheng, Wei-Hong,Kao, Chen-Yi,Hung, Yu-Shin,Su, Po-Jung,Hsieh, Chia-Hsun,Chen, Jen-Shi,Wang, Hung-Ming,Chou, Wen-Chi Asian Pacific Journal of Cancer Prevention 2012 Asian Pacific journal of cancer prevention Vol.13 No.6
Background: The aim of our study was to assess the practical utility of the palliative prognostic index (PPI) as a prognostic tool used by nurse specialists in a hospice consultation setting in Taiwan. Methods: In total, 623 terminal cancer patients under hospice consultation care from one medical center in northern Taiwan were enrolled between January 1 and June 30, 2011. PPI was assessed by a nurse specialist at first hospice consultation and patients categorized into groups by prognosis (good, intermediate, poor). Patient survival was analyzed retrospectively to determine significance of between-group differences. Results: By PPI sum score, 37.2% of patients were in the good prognosis group, 18% in the intermediate prognosis group and 44.8% in the poor prognosis group. The death rates were 56%, 81.2% and 89.6% and median survivals were 76, 18 and 7 days, respectively. The hazard ratio was 0.19 (95% confidence interval [CI] 0.10-0.24, p<0.001) for the poor versus good prognosis group and 0.54 (95% CI 0.43-0.69, p<0.001) for the poor versus intermediate prognosis group. The sensitivity and specificity for the poor prognosis group was 66% and 71%; the positive predictive value and negative predictive value were 81% and 52%, respectively, to predict patient death within 21 days (area under the curve of the receiver operating characteristic was 0.68). Conclusions: Assessment by PPI can accurately predict survival of terminal cancer patients receiving hospice consultation care. PPI is a simple tool and can be administered by nurse members of hospice consultation teams.
Hung-Chun Fu,Jen-Ruei Chen,Min-Yu Chen,Keng-Fu Hsu,Wen-Fang Cheng,An Jen Chiang,Yu-Min Ke,Yu-Chieh Chen,Yin-Yi Chang,Chia-Yen Huang,Chieh-Yi Kang,Yuan-Yee Kan,Sheng-Mou Hsiao,Ming-Shyen Yen 대한부인종양학회 2018 Journal of Gynecologic Oncology Vol.29 No.5
Objective: Choice of hysterectomy and adjuvant treatment for International Federation of Gynecology and Obstetrics (FIGO) 2009 stage II endometrioid endometrial cancer (EEC) is still controversial. Aims of this study were to evaluate survival benefits and adverse effects of different hysterectomies with or without adjuvant radiotherapy (RT), and to identify prognostic factors. Methods: The patients at 14 member hospitals of the Taiwanese Gynecologic Oncology Group from 1992 to 2013 were retrospectively investigated. Patients were divided into simple hysterectomy (SH) alone, SH with RT, radical hysterectomy (RH) alone, and RH with RT groups. Endpoints were recurrence-free survival (RFS), overall survival (OS), disease-specific survival (DSS), adverse effects and prognostic factors for survival. Results: Total of 246 patients were enrolled. The 5-year RFS, OS, DSS and recurrence rates for the entire cohort were 89.5%, 94.3%, 96.2% and 10.2%, respectively. Patients receiving RH had more adverse effects including blood loss (p<0.001), recurrent urinary tract infections (p=0.013), and leg lymphedema (p=0.038). Age over 50-year (HR=9.2; 95% confidence interval [CI], 1.2–70.9) and grade 3 histology (HR=7.28; 95% CI, 1.45–36.6) were independent predictors of OS. Grade 3 histology was an independent predictor of RFS (HR=5.13; 95% CI, 1.38–19.1) and DSS (HR=5.97; 95% CI, 1.06–58.7). Patients receiving adjuvant RT had lower locoregional recurrence (p=0.046), but no impact on survival. Conclusion: Different treatment modalities yield similar survival outcomes. Patients receiving SH with RT had lower locoregional recurrent with acceptable morbidity. Age and tumor grading remained significant predictors for survival among patients with FIGO 2009 stage II EEC.