Filters








13,980 Hits in 3.5 sec

Complement Objective Training [article]

Hao-Yun Chen, Pei-Hsin Wang, Chun-Hao Liu, Shih-Chieh Chang, Jia-Yu Pan, Yu-Ting Chen, Wei Wei, Da-Cheng Juan
2019 arXiv   pre-print
Learning with a primary objective, such as softmax cross entropy for classification and sequence generation, has been the norm for training deep neural networks for years. Although being a widely-adopted approach, using cross entropy as the primary objective exploits mostly the information from the ground-truth class for maximizing data likelihood, and largely ignores information from the complement (incorrect) classes. We argue that, in addition to the primary objective, training also using a
more » ... omplement objective that leverages information from the complement classes can be effective in improving model performance. This motivates us to study a new training paradigm that maximizes the likelihood of the groundtruth class while neutralizing the probabilities of the complement classes. We conduct extensive experiments on multiple tasks ranging from computer vision to natural language understanding. The experimental results confirm that, compared to the conventional training with just one primary objective, training also with the complement objective further improves the performance of the state-of-the-art models across all tasks. In addition to the accuracy improvement, we also show that models trained with both primary and complement objectives are more robust to single-step adversarial attacks.
arXiv:1903.01182v2 fatcat:5vh2g3bt4zgwpfgwwd2jvrctmy

Potent sialic acid inhibitors that target influenza A virus hemagglutinin

Yu-Jen Chang, Cheng-Yun Yeh, Ju-Chien Cheng, Yu-Qi Huang, Kai-Cheng Hsu, Yu-Feng Lin, Chih-Hao Lu
2021 Scientific Reports  
Chang and Cheng-Yun Yeh. * email: chlu@mail.cmu.edu.tw Scientific Reports | (2021) 11:8637 | https://doi.org/10.1038/s41598-021-87845-0 Table 1 . 1 Plaque reduction assay of the nine purchased  ...  Laboratory Science and Biotechnology, Asia University, Taichung, Taiwan. 5 Graduate Institute of Biomedical Sciences, China Medical University, Taichung, Taiwan. 6 These authors contributed equally: Yu-Jen  ... 
doi:10.1038/s41598-021-87845-0 pmid:33883588 pmcid:PMC8060387 fatcat:xfl6liw5izczzbad5mkig6qdui

High Circular Polarized Nanolaser with Chiral Gammadion Metal Cavity

Cheng-Li Yu, Yu-Hao Hsiao, Chiao-Yun Chang, Pi-Ju Cheng, Hsiang-Ting Lin, Ming-Sheng Lai, Hao-Chung Kuo, Shu-Wei Chang, Min-Hsiung Shih
2020 Scientific Reports  
We demonstrate a circularly polarized laser with the metal-gallium-nitride gammadion nanocavities. The ultraviolet lasing signal was observed with the high circular dichroism at room temperature under pulsed optical pump conditions. Without external magnetism which breaks the time-reversal symmetry to favor optical transitions of a chosen handedness, the coherent outputs of these chiral nanolasers show a dissymmetry factor as high as 1.1. The small footprint of these lasers are advantageous for
more » ... applications related to circularly polarized photons in future integrated systems, in contrast to the bulky setup of linearly-polarized lasers and quarter-wave plates.
doi:10.1038/s41598-020-64836-1 pmid:32398835 fatcat:pwkuzl3esfgbdk2hienfmwip6i

Janus Poly(Vinylidene Fluoride) Membranes with Penetrative Pores for Photothermal Desalination

Hao-Hao Yu, Lin-Jiong Yan, Ye-Cheng Shen, Si-Yu Chen, Hao-Nan Li, Jing Yang, Zhi-Kang Xu
2020 Research  
Solar-driven desalination has been considered as a promising technology for producing clean water through an abundant and pollution-free energy source. It is a critical challenge to reasonably design the porous morphology and the thermal management of photothermal membranes for enabling efficient energy conversion and water production. In this work, a Janus poly(vinylidene fluoride) membrane was fabricated in combination of penetrative pore structure, asymmetric surface wettability with proper
more » ... hermal management for high-efficiency solar desalination. Highly open and directly penetrative pores achieved by the two-dimensional solvent freezing strategy are considered to provide direct pathways for water and vapor transportation. The unique feature of hydrophobic upper layer/hydrophilic lower layer enables the photothermal membranes to self-float on the water surface and rapidly pump water from the bulk to the surface. The resulting Janus membrane exhibits a satisfactory light absorbance as high as 97% and a photothermal conversion efficiency of 62.8% under one-sun irradiation in a direct contact mode. The solar-to-vapor efficiency rises up to 90.2% with the assistance of a thermal insulator adopted beneath. Both the Janus membrane and the composite setup are able to work efficiently with a high stability in seawater desalination, and the concentration of ion in condensed water is reduced to below 1 ppm. Therefore, Janus membranes with directly penetrative pores and photothermal surfaces shine a light on the development of high-performance solar evaporators for the practical application in solar seawater desalination.
doi:10.34133/2020/3241758 pmid:32206759 pmcid:PMC7072121 fatcat:2cl4arpbqrdo7kagtgsfo6luse

Self-assembly of polystyrene-b-poly(4-vinylpyridine) in deoxycholic acid melt

Cheng-Hao Yu, Yu-Hao Chuang, Shih-Huang Tung
2011 Polymer  
It is well known that amphiphilic block copolymers in selective solvents self-assemble into micellar structures, where solvophilic blocks tend to contact with solvents while solvophobic blocks are shielded from the solvents. Different from the conventional micellization in liquid systems, we report that the block copolymer, polystyrene-b-(4-vinylpyridine) (PS-b-P4VP), can self-assemble in melted deoxycholic acid (DCA) at high temperatures and the structures are retained in "solid state" after
more » ... ing cooled down to room temperature. Probing by transmission electron microscopy (TEM), we found that a series of selfassembled structures, including spherical micelles, wormlike micelles and vesicles can be obtained by varying the length of the block copolymers and the morphologies are dependent on the annealing temperature and time. We also demonstrate how to extract the structures that are trapped in solid state by removing DCA using appropriate solvents. The extracted vesicles, which are loaded with solid molecules, are potential for applications in nanocapsules and controlled release.
doi:10.1016/j.polymer.2011.07.014 fatcat:epy7us5tfbdcffokdjfx6ao2em

Fast Distributed Complex Join Processing [article]

Hao Zhang, Miao Qiao, Jeffrey Xu Yu, Hong Cheng
2021 arXiv   pre-print
In this work, we study the problem of co-optimize communication, pre-computing, and computation cost in one-round multi-way join evaluation. We propose a multi-way join approach ADJ (Adaptive Distributed Join) for complex join which finds one optimal query plan to process by exploring cost-effective partial results in terms of the trade-off between pre-computing, communication, and computation.We analyze the input relations for a given join query and find one optimal over a set of query plans
more » ... some specific form, with high-quality cost estimation by sampling. Our extensive experiments confirm that ADJ outperforms the existing multi-way join methods by up to orders of magnitude.
arXiv:2102.13370v1 fatcat:anq56b4uonfghh7lxhpzmaxbpq

Prim-SNPing: a primer designer for cost-effective SNP genotyping

Hsueh-Wei Chang, Li-Yeh Chuang, Yu-Huei Cheng, Yu-Chen Hung, Cheng-Hao Wen, De-Leung Gu, Cheng-Hong Yang
2009 BioTechniques  
doi:10.2144/000113092 pmid:19480636 fatcat:477v3kuhnbbdrdlhux5dtalm4q

Effective string processing and matching for author disambiguation

Wei-Sheng Chin, Wei-Cheng Chang, Kuan-Hao Huang, Tzu-Ming Kuo, Shan-Wei Lin, Young-San Lin, Yu-Chen Lu, Yu-Chuan Su, Cheng-Kuang Wei, Tu-Chun Yin, Chun-Liang Li, Yu-Chin Juan (+12 others)
2013 Proceedings of the 2013 KDD Cup 2013 Workshop on - KDD Cup '13  
Track 2 in KDD Cup 2013 aims at determining duplicated authors in a data set from Microsoft Academic Search. This type of problems appears in many large-scale applications that compile information from different sources. This paper describes our solution developed at National Taiwan University to win the first prize of the competition. We propose an effective name matching framework and realize two implementations. An important strategy in our approach is to consider Chinese and non-Chinese
more » ... s separately because of their different naming conventions. Post-processing including merging results of two predictions further boosts the performance. Our approach achieves F1-score 0.99202 on the private leader board, while 0.99195 on the public leader board.
doi:10.1145/2517288.2517295 dblp:conf/kdd/ChinJZWTYWCYCHK13 fatcat:d36i5rr2pbeothwm4267ccjqwm

Measuring the Sustainability of Construction Projects throughout Their Lifecycle: A Taiwan Lesson

Wen-der Yu, Shao-tsai Cheng, Wei-cheng Ho, Yu-hao Chang
2018 Sustainability  
Wei-cheng Ho and Yu-hao Chang, who were graduate students of Yu and Cheng, implemented the research methodology and conducted questionnaire survey.  ...  The statistical analysis and validation was conducted by Cheng.  ... 
doi:10.3390/su10051523 fatcat:7rjbelgas5e7vnhtgpwkjs5lta

The rapid effects of eicosapentaenoic acid (EPA) enriched phospholipids on alleviating exercise fatigue in mice

Cheng-Cheng Wang, Hao-Hao Shi, Ling-Yu Zhang, Lin Ding, Chang-Hu Xue, Teruyoshi Yanagita, Tian-Tian Zhang, Yu-Ming Wang
2019 RSC Advances  
EPA-PL has rapid effects on alleviating exercise fatigue by inhibiting lactic acid accumulation, glycogen consumption and muscle injury.
doi:10.1039/c9ra05181c pmid:35528913 pmcid:PMC9073713 fatcat:qiqc4bav3zecnpwbour7rrlq4e

Smoking Status and Functional Outcomes in Young Stroke

Cheng-Loong Liang, Han-Jung Chen, Yi-Che Lee, Cheng-Chun Wu, Chon-Haw Tsai, Po-Lin Chen, Wei-Lun Chang, Po-Yen Yeh, Cheng-Yu Wei, Ming-Jun Tsai, Yu Sun, Chih-Hao Lin (+7 others)
2021 Frontiers in Neurology  
Objective: Stroke in young adults is uncommon, and the etiologies and risk factors of stroke in young adults differ from those in older populations. Smoker's paradox is an unexpected favorable outcome, and age difference is used to explain the association between smoking and the favorable functional outcome. This study aimed to investigate the existence of this phenomenon in young stroke patients.Methods: We analyzed a total of 9,087 young stroke cases registered in the nationwide stroke
more » ... y system of Taiwan between 2006 and 2016. Smoking criteria included having a current history of smoking more than one cigarette per day for more than 6 months. After matching for sex and age, a Cox model was used to compare mortality and function outcomes between smokers and non-smokers.Results: Compared with the non-smoker group, smoking was associated with older age, higher comorbidities, and higher alcohol consumption. Patients who report smoking with National Institutes of Health Stroke Scale scores of 11–15 had a worse functional outcome (adjusted odds ratio, 0.81; 95% confidence interval, 0.76 – 0.87).Conclusion: Smokers had a higher risk of unfavorable functional outcomes at 3 months after stroke, and therefore, we continue to strongly advocate the importance of smoking cessation.
doi:10.3389/fneur.2021.658582 pmid:34539539 pmcid:PMC8440842 fatcat:6drpt537fvbf3g3saye73xxxzy

Parallelism-Oriented Dynamic Incremental Delaunay Triangulation Algorithm

YANG Haoyu, LIU Li, ZHANG Cheng, YU Hao
2020 Jisuanji kexue yu tansuo  
Delaunay triangulation is a main topic in computer graphics. Various types of new requirements have appeared during the development of parallel triangulation algorithms, e.g., updating the triangulation incrementally given a set of increasing points. Existing incremental triangulation algorithms do not support for inserting points outside of the triangulation. This paper proposes the expansion of triangulation and designs an incremental triangulation algorithm TID (truly incremental Delaunay)
more » ... pporting for inserting any number of points with any distribution into an existing triangulation for any time. TID is designed to produce a unique triangulation result for any distribution of points. Experimental evaluation results on TID show that TID has higher computational efficiency than existing algorithms and introduces low computational overhead. In addition, TID has already been used in parallel triangulation algorithm as local triangulation algorithm.
doi:10.3778/j.issn.1673-9418.1903014 doaj:4bd8b162ea86423a97d2ac47e957b68b fatcat:dfrzcequ55dntlwfrrra4szfl4

Preoperative Inversed Albumin-to-Globulin Ratio Predicts Worse Oncologic Prognosis Following Curative Hepatectomy for Hepatocellular Carcinoma

Cheng-Cheng Zhang, Cheng-Wu Zhang, Hao Xing, Yu Wang, Lei Liang, Yong-Kang Diao, Ting-Hao Chen, Wan Yee Lau, Ping Bie, Zhi-Yu Chen, Tian Yang
2020 Cancer Management and Research  
A normal albumin-to-globulin ratio (NAGR) in serum is greater than 1. Inversed albumin-to-globulin ratio (IAGR < 1) indicates poor synthetic liver function or malnutrition. The aim of this study is to evaluate whether preoperative IAGR was associated with worse oncologic survival after hepatectomy for hepatocellular carcinoma (HCC). Patients who underwent curative hepatectomy for HCC between 2009 and 2016 in four centers were divided into the IAGR and NAGR groups based on their preoperative
more » ... ls, and their clinical characteristics and long-term survival outcomes were compared. Univariable and multivariable Cox regression analyses were performed to identify risk factors of overall survival (OS) and recurrence-free survival (RFS). Of 693 enrolled patients, 136 (19.6%) were in the IAGR group. Their 5-year OS and RFS rates were 31.6% and 21.3%, respectively, which were significantly worse than the NAGR group (43.4% and 28.7%, both P < 0.001). The area under the receiver operating characteristic curves in predicting 5-year OS and RFS using the albumin-to-globulin ratio were 0.68 and 0.67, respectively, which were significantly higher than albumin (0.60 and 0.59), globulin (0.56 and 0.57), Child-Pugh grading (0.61 and 0.60), Model for End-Stage Liver Disease Score (0.59 and 0.58), and Albumin-Bilirubin grading (0.64 and 0.63). Multivariable analyses identified that preoperative IAGR was independently associated with worse OS (HR: 1.444, 95% confidence interval (CI): 1.125-1.854, P = 0.004) and RFS (HR: 1.463, 95% CI: 1.159-1.848, P = 0.001). Preoperative IAGR was useful in predicting worse OS and RFS in patients who underwent curative hepatectomy for HCC.
doi:10.2147/cmar.s275307 pmid:33116850 pmcid:PMC7567562 fatcat:v7n6cerfb5e5lgsism53lz3lhm

Residual Bootstrap Exploration for Bandit Algorithms [article]

Chi-Hua Wang, Yang Yu, Botao Hao, Guang Cheng
2020 arXiv   pre-print
Correspondence to: Guang Cheng <chengg@purdue.edu>. are based on the optimism principle.  ...  Hao et al. (2019) proposed a nonparametric and data-dependent UCB algorithm based on the multiplier bootstrap and derived a second-order correction term to boost the agent away from sub-optimal solutions  ... 
arXiv:2002.08436v1 fatcat:5vxbishzw5amfiknc3extet3bm

Learning with Hierarchical Complement Objective [article]

Hao-Yun Chen, Li-Huang Tsai, Shih-Chieh Chang, Jia-Yu Pan, Yu-Ting Chen, Wei Wei, Da-Cheng Juan
2019 arXiv   pre-print
Label hierarchies widely exist in many vision-related problems, ranging from explicit label hierarchies existed in image classification to latent label hierarchies existed in semantic segmentation. Nevertheless, state-of-the-art methods often deploy cross-entropy loss that implicitly assumes class labels to be exclusive and thus independence from each other. Motivated by the fact that classes from the same parental category usually share certain similarity, we design a new training diagram
more » ... d Hierarchical Complement Objective Training (HCOT) that leverages the information from label hierarchy. HCOT maximizes the probability of the ground truth class, and at the same time, neutralizes the probabilities of rest of the classes in a hierarchical fashion, making the model take advantage of the label hierarchy explicitly. The proposed HCOT is evaluated on both image classification and semantic segmentation tasks. Experimental results confirm that HCOT outperforms state-of-the-art models in CIFAR-100, ImageNet-2012, and PASCAL-Context. The study further demonstrates that HCOT can be applied on tasks with latent label hierarchies, which is a common characteristic in many machine learning tasks.
arXiv:1911.07257v1 fatcat:slb5yavfpbeihnuwxxb6rg2ybq
« Previous Showing results 1 — 15 out of 13,980 results