335 Hits in 1.5 sec

Local Reweighting for Adversarial Training [article]

Ruize Gao, Feng Liu, Kaiwen Zhou, Gang Niu, Bo Han, James Cheng
2021 arXiv   pre-print
Instances-reweighted adversarial training (IRAT) can significantly boost the robustness of trained models, where data being less/more vulnerable to the given attack are assigned smaller/larger weights during training. However, when tested on attacks different from the given attack simulated in training, the robustness may drop significantly (e.g., even worse than no reweighting). In this paper, we study this problem and propose our solution--locally reweighted adversarial training (LRAT). The
more » ... tionale behind IRAT is that we do not need to pay much attention to an instance that is already safe under the attack. We argue that the safeness should be attack-dependent, so that for the same instance, its weight can change given different attacks based on the same model. Thus, if the attack simulated in training is mis-specified, the weights of IRAT are misleading. To this end, LRAT pairs each instance with its adversarial variants and performs local reweighting inside each pair, while performing no global reweighting--the rationale is to fit the instance itself if it is immune to the attack, but not to skip the pair, in order to passively defend different attacks in future. Experiments show that LRAT works better than both IRAT (i.e., global reweighting) and the standard AT (i.e., no reweighting) when trained with an attack and tested on different attacks.
arXiv:2106.15776v1 fatcat:o3vndlnw2far5a7hsdnpxp2l2y

Mutagenesis facilitated crystallization of GLP-1R

Yueming Xu, Yuxia Wang, Yang Wang, Kaiwen Liu, Yao Peng, Deqiang Yao, Houchao Tao, Haiguang Liu, Gaojie Song
2019 IUCrJ  
Stevens and Professor Zhi-Jie Liu for their mentoring and assistance on this project. Funding information Xu et al.  ... 
doi:10.1107/s2052252519013496 pmid:31709055 pmcid:PMC6830218 fatcat:sakulrng4ben7ctnruwia3fxo4

Deep Reinforcement Learning for Orienteering Problems Based on Decomposition [article]

Wei Liu, Tao Zhang, Rui Wang, Kaiwen Li, Wenhua Li, Kang Yang
2022 arXiv   pre-print
Wei Liu, Tao Zhang, Rui Wang, Kaiwen Li, Wenhua Li and Kang Yang are with the College of System Engineering, National University of Defense Technology, Changsha 410073, PR China, and with the Hunan Key  ... 
arXiv:2204.11575v1 fatcat:hplcreteivdx5l4pk3llrbpb2e

Anti-parity-time topologically undefined state [article]

Haohao Wang, Kaiwen Ji, Yuandan Wang, Zhenjuan Liu, Yuanmei Gao, Yanlong Shen, Shi Bai, Koji Sugioka, Xinyuan Qi
2021 arXiv   pre-print
We constructed an anti-parity-time-symmetric photonic lattice by using perturbations. The results show the topological state will appear when the waveguide coupling constants κ_1<κ_2; Interestingly, a state with undefined winding numbers occurs when κ_1=κ_2, in which the light distributes only in the wide waveguides with equal magnitude distribution. Further studies show that the edge state will be strengthened by introducing defect for the topologically non-trivial case, while it will not
more » ... t the equal intensity transmission for the topologically undefined state. Our work provides a new way to realize the topological state and equally divided light transmission and might be applicable in optical circuits and optical interconnect.
arXiv:2107.02533v1 fatcat:pgy4ce727nh2dcxa6b3vm4zmzy

Guaranteed Sufficient Decrease for Stochastic Variance Reduced Gradient Optimization [article]

Fanhua Shang, Yuanyuan Liu, Kaiwen Zhou, James Cheng, Kelvin K.W. Ng, Yuichi Yoshida
2018 arXiv   pre-print
Supplementary Materials for "Guaranteed Sufficient Decrease for Stochastic Variance Reduced Gradient Optimization" Fanhua Shang, Yuanyuan Liu, Kaiwen Zhou, James Cheng, Kelvin K.W.  ...  Liu, F. Shang, and J. Cheng. Accelerated vari- ance reduced stochastic ADMM. In AAAI, pages 2287-2293, 2017. [31] Y. Nesterov.  ... 
arXiv:1802.09933v1 fatcat:l2umcy4oyffqxf3fyy5zm5ey6y

Deep learning methods improve linear B-cell epitope prediction

Tao Liu, Kaiwen Shi, Wuju Li
2020 BioData Mining  
51.22 51.03 50.61 52.18 51.4 50.32 55.79 61.88 Bepipred2.0 51.89 52.10 51.26 52.8 54.02 53.85 53.23 55.33 DLBEpitope 88.80 89.98 90.85 91.58 93.95 95.68 73.83 65.5 Liu  ... 
doi:10.1186/s13040-020-00211-0 pmid:32699555 pmcid:PMC7371472 fatcat:cjbjjzgdrzhnllrj3mxvtwyj7a

A Proteomic View of Salmonella Typhimurium in Response to Phosphate Limitation

Jiezhang Jiang, Kaiwen Yu, Linlu Qi, Yanhua Liu, Sen Cheng, Mei Wu, Zhen Wang, Jiaqi Fu, Xiaoyun Liu
2018 Proteomes  
Acknowledgments: We thank the members of the Liu laboratory for critical reading of this manuscript. Conflicts of Interest: The authors declare no conflict of interest.  ... 
doi:10.3390/proteomes6020019 pmid:29693629 pmcid:PMC6027262 fatcat:7xd4rvn2mzaxtcmmbhf5haizgm

CoDiM: Learning with Noisy Labels via Contrastive Semi-Supervised Learning [article]

Xin Zhang, Zixuan Liu, Kaiwen Xiao, Tian Shen, Junzhou Huang, Wei Yang, Dimitris Samaras, Xiao Han
2021 arXiv   pre-print
C2D (Zheltonozhskii et al. 2021) evaluates its effectiveness to initialize the model for different LNL methods such as DivideMix and ELR+ (Liu et al. 2020) .  ...  criterions such as small-loss (Jiang et al. 2018; Han et al. 2018) and prediction disagreement (Malach and Shalev-Shwartz 2017) , or directly applying regularization through early-stop strategy (Liu  ... 
arXiv:2111.11652v1 fatcat:4qfk4z4eazbxviswhwjb5lvawy

Progress in Animal Models of Pancreatic Ductal Adenocarcinoma

Kaiwen Kong, Meng Guo, Yanfang Liu, Jianming Zheng
2020 Journal of Cancer  
As a common gastrointestinal tumor, the incidence of pancreatic cancer has been increasing in recent years. The disease shows multi-gene, multi-step complex evolution from occurrence to dissemination. Furthermore, pancreatic cancer has an insidious onset and an extremely poor prognosis, so it is difficult to obtain cinical specimens at different stages of the disease, and it is, therefore, difficult to observe tumorigenesis and tumor development in patients with pancreatic cancer. At present,
more » ... standard protocols stipulate clinical treatment of pancreatic cancer, and the benefit rate of new targeted therapies is low. For this reason, a well-established preclinical model of pancreatic cancer must be established to allow further exploration of the occurrence, development, invasion, and metastasis mechanism of pancreatic cancer, as well as to facilitate research into new therapeutic targets. A large number of animal models of pancreatic cancer are currently available, including a cancer cell line-based xenograft, a patient-derived xenograft, several mouse models (including transgenic mice), and organoid models. These models have their own characteristics, but they still cannot perfectly predict the clinical outcome of the new treatment. In this paper, we present the distinctive features of the currently popular pancreatic cancer models, and discuss their preparation methods, clinical relations, scientific purposes and limitations.
doi:10.7150/jca.37529 pmid:32047562 pmcid:PMC6995380 fatcat:qqb5gjmruzd2llemmlcfdc2v5q

Identification of a NovelSalmonellaType III Effector by Quantitative Secretome Profiling

Sen Cheng, Lu Wang, Qian Liu, Linlu Qi, Kaiwen Yu, Zhen Wang, Mei Wu, Yanhua Liu, Jiaqi Fu, Mo Hu, Min Li, Daoguo Zhou (+1 others)
2017 Molecular & Cellular Proteomics  
Acknowledgments-We thank the members of the Liu laboratory for careful review of this manuscript. We also thank Dr. Xing Chen for the use of the Zeiss LSM700 confocal microscope and Dr.  ... 
doi:10.1074/mcp.ra117.000230 pmid:28887382 fatcat:i3dysdtqfzexhdmrd3yn6n263e

Effects of Diabetes on Inflammatory Status and Prognosis in Cancer Patients

Xiangliang Liu, Kaiwen Zheng, Wei Ji, Wenxin Zhang, Yuguang Li, Mingyang Liu, Jiuwei Cui, Wei Li
2022 Frontiers in Nutrition  
BackgroundCancer and diabetes mellitus (DM) are prevalent, but there still a lack of convinced evidence clearly explaining the extent of the effect of diabetes in cancer.Data and MethodsClinical data of 2,929 cancer patients were collected. Diabetes were diagnosed according to the Diabetes Diagnosis and Treatment Criteria. BMI was classified by the BMI standards for Chinese adults published by the Working Group on Obesity. All involved patients were classified into the non-DM group and DM
more » ... The Kaplan–Meier curve, log-rank test and Cox regression analyses were used to perform survival analysis.ResultsCompared with non-DM patients, OS in DM patients was significant shorter in lung cancer (HR = 2.076, P = 0.001 in early stage; HR = 2.118, P &lt; 0.001 in advanced stage), digestive tract cancer (HR = 1.768, P = 0.020 in early stage; HR = 2.454, P = 0.005 in advanced stage), leukemia (HR = 2.636, P &lt; 0.001), breast cancer (HR = 2.495, P = 0.047 in early stage; HR = 2.929, P = 0.019 in advanced stage) and liver cancer (HR = 3.086, P &lt; 0.001 in early stage; HR = 2.219, P = 0.049 in advanced stage). DM negatively influenced OS when the BMI was within the normal range in overall cancer (HR = 2.468, P &lt; 0.001), lung cancer (HR = 2.297, P &lt; 0.001), digestive tract cancer (HR = 2.354, P &lt; 0.001), liver cancer (HR = 2.406, P = 0.001), leukemia (HR = 4.039, P &lt; 0.001) and breast cancer (HR = 4.222, P = 0.008). Among those with BMI ≥ 24 kg/m2, DM played a role only in lung cancer (HR = 1.597, P = 0.037).ConclusionsPatients with diabetes tend to combine worse body composition and inflammation status, and that glycemic control can ameliorate the impairment of diabetes to some extent.
doi:10.3389/fnut.2022.792577 pmid:35308284 pmcid:PMC8931397 fatcat:erosxuuu45aezmauarvfu3odce

Multi-task Learning for Macromolecule Classification, Segmentation and Coarse Structural Recovery in Cryo-Tomography [article]

Chang Liu, Xiangrui Zeng, Kaiwen Wang, Qiang Guo, Min Xu
2018 arXiv   pre-print
In particular, Liu et al [15] have proposed a 3DCNN based segmentation network, named SSN3D-ED which is capable of masking out neighboring structures inside subtomograms while maintaining generalization  ... 
arXiv:1805.06332v1 fatcat:wqluw4qpprhilhed43ulir4kn4

An alternative route towards low-cost Cu2ZnSnS4 thin film solar cells

Ding Tang, Qinglong Wang, Fangyang Liu, Lianbo Zhao, Zili Han, Kaiwen Sun, Yanqing Lai, Jie Li, Yexiang Liu
2013 Surface & Coatings Technology  
A non-vacuum and low-cost route to fabrication of Cu 2 ZnSnS 4 (CZTS) thin film as absorber layer for thin film solar cell is demonstrated. The process involves synthesis of Cu-Zn-Sn composite oxide nanoparticles, preparation of composite oxide ink, deposition of composite oxide thin film and subsequent sulfurization treatment. Cu-Zn-Sn composite oxide nanoparticles with proper composition and particle size can be obtained by controlling the chemical co-precipitation conditions. The effects of
more » ... ulfurization temperature on the compositional, structural, morphological, optical and electrical properties of the CZTS thin films were investigated. XRD, Raman and X-ray photoelectron spectroscopy analyses confirmed the formation of CZTS phase after sulfurization. The prepared CZTS thin films exhibit porous morphology and an optical band gap about 1.35 eV. The solar cells fabricated with the CZTS thin film sulfurized at 550°C showed a conversion efficiency of 1.22%.
doi:10.1016/j.surfcoat.2013.04.052 fatcat:g5dovw6lgfharjreeprslssyvi

Research on Operation Intention Based on Flexible Tactile Sensing Handle

Tiejun Li, Kaiwen Zheng, Jinyue Liu, Xiaohui Jia, Jianbin Feng
2021 IEEE Access  
In order to obtain and analyze the operator's intention comprehensively and accurately in human-robot interaction, an array-type flexible tactile sensor was designed. The sensor was encapsulated into a tactile handle to sense the grasping state of the human hand in real time. According to the analysis of different operators' grasping posture and grasping habits, the grasping state was defined as 5 modes. Based on Harris feature point positioning and extraction, a method of grasping posture
more » ... rsion was proposed to ensure the completeness and standard of the extracted grasping features. A set of Convolutional Neural Networks (CNN) suitable for the real-time classification of the grasping intention was built to distinguish the grasping state sensed by the handle in real time, accurately determine the operator's intention, and complete the interaction with the robot. Using a UR collaborative robot as the experimental platform and the haptic handle as the intent sensing device, the intent-behaviour mapping relationship was constructed to control the motion of the UR collaborative robot. The experimental results show that the classification accuracy of operation intention is as high as 97.87%.
doi:10.1109/access.2021.3050991 fatcat:66j34gheljcfzhd532bof23bly

MiR-449a regulates caprine endometrial stromal cell apoptosis and endometrial receptivity

Xiaopeng An, Xiaorui Liu, Lei Zhang, Junze Liu, Xinyan Zhao, Kaiwen Chen, Haidong Ma, Guang Li, Binyun Cao, Yuxuan Song
2017 Scientific Reports  
In this study, an RT-qPCR analysis showed that the expression levels of miR-449a in the pre-receptive endometrium were lower compared to the receptive endometrium, which is consistent with previous sequencing data (previous investigations). To detect the role of miR-449a in endometrial receptivity, we transfected caprine endometrial stromal cells (ESCs) cultured in vitro with miR-449a mimics. The results revealed that miR-449a decreased the mRNA and protein levels of LGR4 by targeting its
more » ... ranslated region. The miR-449a mimics significantly reduced the G 1 cell population from 52.56% (mimic NC) to 42.19% with a concordant increase in the G 2 and S cell populations from 47.44% (mimic NC) to 57.81%, suggesting that miR-449a caused ESC cell cycle arrest. In addition, the number of apoptotic cells was significantly higher in ESCs transfected with miR-449a mimics (P < 0.05) than in ESCs transfected with mimic NC. In vivo, rich pinopodes were observed on the endometrial surface in the miR-449a agomir group compared with the miR-449a antagomir group. The results of hematoxylin-eosin staining showed that endometrial thickness was significantly increased in the miR-449a agomir group compared with the miR-449a antagomir group. These results suggest that miR-449a could enhance endometrial receptivity. Published: xx xx xxxx OPEN 2 Scientific RepoRts | 7: 12248 |
doi:10.1038/s41598-017-12451-y pmid:28947781 pmcid:PMC5612931 fatcat:sz4uvwfduvegzotkr5d5j4v5ku
« Previous Showing results 1 — 15 out of 335 results