http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Optimizing TensorFlow Performance by Reconstructing the Convolution Routine
Minseong Kim,Kyu Hyun Choi,Yoonah Paik,Seon Wook Kim 대한전자공학회 2021 IEIE Transactions on Smart Processing & Computing Vol.10 No.2
Using deep learning, we can currently build computational models composed of multiple processing layers to learn representations of data. Convolutional neural networks (CNNs) have been widely adopted to achieve significant performance in image recognition and classification. TensorFlow, an open-source deep learning framework from Google, uses profiling to select one convolution algorithm, from among several available, as the core of a CNN to deliver the best performance in terms of execution time and memory usage. However, the overhead from profiling is considerably significant, because TensorFlow executes and profiles all the available algorithms for the best selection whenever an application is launched. We observe that memory usage overshoots during profiling, which limits data parallelism, and thus, fails to deliver maximum performance. In this paper, we present a novel profiling method to reduce overhead by storing the profile result from the first run and reusing it from the second run on. Using Inception-V3, we achieved up to 1.12 times and 1.11 times higher throughput, compared to the vanilla TensorFlow and TensorFlow with XLA JIT compilation, respectively, without losing accuracy.
Corporate Social Responsibility: A Comparison Analysis
Yoonah Hahn,Dongho Kim 한국유통과학회 2016 Asian Journal of Business Environment (AJBE) Vol.6 No.4
Purpose - The purpose of this paper is to evaluate two multinational companies that seem to have reconciled the two mandates of CSR and profit maximization while becoming multibillion dollar companies and examine their organizational culture and practices and their management and leadership in order to determine the controlling factors, if any, that have elicited their success while renowned for their CSR policies. Research design, data, and methodology - This is a case study, an analytical approach, which focuses on exploring and analyzing the CSR policies of Starbucks and IKEA. Results – IKEA and Starbucks considered their position in the global business environment and their social responsibilities as crucial and did more than a cursory lip service to the issues. In fact, they both took the more difficult long-term approach and tried to resolve the root causes for the environmental and social issues in their supply chain. Ultimately though, it is the ethical leadership of the top management that sets the tone for the organizational culture and its CSR. Conclusion - IKEA and Starbucks are “living proof” that a company can be successful while treating its employees and the community of suppliers and associates with respect and dignity and while making this world a better place.
Learning curves for single incision and conventional laparoscopic right hemicolectomy
Yoonah Park,Yuen Geng Yong,Seong Hyeon Yun,Kyung Uk Jung,Jung Wook Huh,Yong Beom Cho,Hee Cheol Kim,Woo Yong Lee,Ho-Kyung Chun 대한외과학회 2015 Annals of Surgical Treatment and Research(ASRT) Vol.88 No.5
Purpose: This study aimed to compare the learning curves and early postoperative outcomes for conventional laparoscopic (CL) and single incision laparoscopic (SIL) right hemicolectomy (RHC). Methods: This retrospective study included the initial 35 cases in each group. Learning curves were evaluated by the moving average of operative time, mean operative time of every five consecutive cases, and cumulative sum (CUSUM) analysis. The learning phase was considered overcome when the moving average of operative times reached a plateau, and when the mean operative time of every five consecutive cases reached a low point and subsequently did not vary by more than 30 minutes. Results: Six patients with missing data in the CL RHC group were excluded from the analyses. According to the mean operative time of every five consecutive cases, learning phase of SIL and CL RHC was completed between 26 and 30 cases, and 16 and 20 cases, respectively. Moving average analysis revealed that approximately 31 (SIL) and 25 (CL) cases were needed to complete the learning phase, respectively. CUSUM analysis demonstrated that 10 (SIL) and two (CL) cases were required to reach a steady state of complication-free performance, respectively. Postoperative complications rate was higher in SIL than in CL group, but the difference was not statistically significant (17.1% vs. 3.4%). Conclusion: The learning phase of SIL RHC is longer than that of CL RHC. Early oncological outcomes of both techniques were comparable. However, SIL RHC had a statistically insignificant higher complication rate than CL RHC during the learning phase.