Search

Your search keyword '"Peng, Liwei"' showing total 139 results
Start Over Author "Peng, Liwei"
139 results on '"Peng, Liwei"'

7. StructBERT: Incorporating Language Structures into Pre-training for Deep Language Understanding

Author

Wang, Wei, Bi, Bin, Yan, Ming, Wu, Chen, Bao, Zuyi, Xia, Jiangnan, Peng, Liwei, and Si, Luo

Subjects
Computer Science - Computation and Language
Abstract

Recently, the pre-trained language model, BERT (and its robustly optimized version RoBERTa), has attracted a lot of attention in natural language understanding (NLU), and achieved state-of-the-art accuracy in various NLU tasks, such as sentiment classification, natural language inference, semantic textual similarity and question answering. Inspired by the linearization exploration work of Elman [8], we extend BERT to a new model, StructBERT, by incorporating language structures into pre-training. Specifically, we pre-train StructBERT with two auxiliary tasks to make the most of the sequential order of words and sentences, which leverage language structures at the word and sentence levels, respectively. As a result, the new model is adapted to different levels of language understanding required by downstream tasks. The StructBERT with structural pre-training gives surprisingly good empirical results on a variety of downstream tasks, including pushing the state-of-the-art on the GLUE benchmark to 89.0 (outperforming all published models), the F1 score on SQuAD v1.1 question answering to 93.0, the accuracy on SNLI to 91.7., Comment: 10 Pages

Published
2019

10. The comparison of the FRF curvature technique and FRF mutation coefficient technique on the identification effect of beam structure damage location.

Author

Guo, Tieneng, Zhou, Cheng, Peng, Liwei, and Meng, Lingjun

Abstract

Improving the rigidity of machine tools for identifying the weak links of structural stiffness is one of the important issues in mechanical engineering. In this paper, a new weak index is proposed on a cantilever structure of the machine tool to identify the weak stiffness link. The weak index is a frequency parameter obtained via the vibration test method. In this method, the FRF coefficient of variation technique is proposed based on the FRF curvature method. The change of weak index is used to identify the weak link in the structural stiffness. The element stiffness weakening method that employs numerical examples of a cantilever beam and an elastic support beam is used to simulate the weak links of the stiffness. Compared with the FRF curvature ratio (FRF_CR) method, the FRF mutation coefficient is improved in the same frequency range. Moreover, the FRF_MCR indicator has more frequency points than the FRF_CR indicator in the same frequency range. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

15. The comparison of the FRF curvature technique and FRF mutation coefficient technique on the identification effect of beam structure damage location

Author

da Silva, Lucas FM, Adams, Robert D, Sato, Chiaki, Dilger, Klaus, Guo, Tieneng, Zhou, Cheng, Peng, Liwei, and Meng, Lingjun

Abstract

Improving the rigidity of machine tools for identifying the weak links of structural stiffness is one of the important issues in mechanical engineering. In this paper, a new weak index is proposed on a cantilever structure of the machine tool to identify the weak stiffness link. The weak index is a frequency parameter obtained via the vibration test method. In this method, the FRF coefficient of variation technique is proposed based on the FRF curvature method. The change of weak index is used to identify the weak link in the structural stiffness. The element stiffness weakening method that employs numerical examples of a cantilever beam and an elastic support beam is used to simulate the weak links of the stiffness. Compared with the FRF curvature ratio (FRF_CR) method, the FRF mutation coefficient is improved in the same frequency range. Moreover, the FRF_MCR indicator has more frequency points than the FRF_CR indicator in the same frequency range.

Published
2024
Full Text
View/download PDF

24. Alcohol Abuse Associated With Increased Risk of Angiographic Vasospasm and Delayed Cerebral Ischemia in Patients With Aneurysmal Subarachnoid Hemorrhage Requiring Mechanical Ventilation

Author

Zhao, Lei, primary, Cheng, Chao, additional, Peng, Liwei, additional, Zuo, Wei, additional, Xiong, Dong, additional, Zhang, Lei, additional, Mao, Zilong, additional, Zhang, Jin'an, additional, Wu, Xia, additional, Jiang, Xue, additional, Wang, Peng, additional, and Li, Weixin, additional

Published
2022
Full Text
View/download PDF

27. Additional file 1 of Machine learning approach for the prediction of 30-day mortality in patients with sepsis-associated encephalopathy

Author

Peng, Liwei, Peng, Chi, Yang, Fan, Wang, Jian, Zuo, Wei, Cheng, Chao, Mao, Zilong, Jin, Zhichao, and Li, Weixin

Subjects
ComputingMethodologies_PATTERNRECOGNITION
Abstract

Additional file 1: Table S1. Missing number (%) for included variables in the dataset. Table S2. The comparison of different ROC curve by De Long’s method.

Published
2022
Full Text
View/download PDF

34. Suppression of LjBAK1-mediated immunity by SymRK promotes rhizobial infection in Lotus japonicus

Author

Feng, Yong, primary, Wu, Ping, additional, Liu, Chao, additional, Peng, Liwei, additional, Wang, Tao, additional, Wang, Chao, additional, Tan, Qian, additional, Li, Bixuan, additional, Ou, Yajuan, additional, Zhu, Hui, additional, Yuan, Songli, additional, Huang, Renliang, additional, Stacey, Gary, additional, Zhang, Zhongming, additional, and Cao, Yangrong, additional

Published
2021
Full Text
View/download PDF

41. EFLOPS: Algorithm and system co-design for a high performance distributed training platform

Author

Dong, Jianbo, Cao, Zheng, Zhang, Tao, Ye, Jianxi, Wang, Shaochuang, Feng, Fei, Zhao, Li, Liu, Xiaoyong, Song, Liuyihan, Peng, Liwei, Guo, Yiqun, Jiang, Xiaowei, Tang, Lingbo, Du, Yin, Zhang, Yingya, Pan, Pan, Xie, Yuan, Dong, Jianbo, Cao, Zheng, Zhang, Tao, Ye, Jianxi, Wang, Shaochuang, Feng, Fei, Zhao, Li, Liu, Xiaoyong, Song, Liuyihan, Peng, Liwei, Guo, Yiqun, Jiang, Xiaowei, Tang, Lingbo, Du, Yin, Zhang, Yingya, Pan, Pan, and Xie, Yuan

Abstract

Deep neural networks (DNNs) have gained tremendous attractions as compelling solutions for applications such as image classification, object detection, speech recognition, and so forth. Its great success comes with excessive trainings to make sure the model accuracy is good enough for those applications. Nowadays, it becomes challenging to train a DNN model because of 1) the model size and data size keep increasing, which usually needs more iterations to train; 2) DNN algorithms evolve rapidly, which requires the training phase to be short for a quick deployment. To address those challenges, distributed training platforms have been proposed to leverage massive server nodes for training, with the hope of significant training time reduction. Therefore, scalability is a critical performance metric to evaluate a distributed training platform. Nevertheless, our analysis reveals that traditional server clusters have poor scalability for training due to the traffic congestions within the server and beyond. The intra-server traffic on the I/O fabric can result in severe congestions and skewed quality of service as high performance devices are competing with each other. Moreover, the traffic congestions on the Ethernet for inter-server communication could also incur significant performance degradation. In this work, we devise a novel distributed training platform, EFLOPS, that adopts an algorithm and system co-design methodology to achieve good scalability. A new server architecture is proposed to alleviate the intra-server congestions. Moreover, a new network topology, BiGraph, is proposed to divide the network into two separate parts, so that there is always a direct connection between any nodes from different parts. Finally, accompany with BiGraph, a topology-aware allreduce algorithm is proposed to eliminate the traffic congestion on the direct connection. The experimental results show that eliminating the congestions on network interface can gain up to 11.3xcommunicatio

Published
2020

48. EFLOPS: Algorithm and System Co-Design for a High Performance Distributed Training Platform

Author

Dong, Jianbo, primary, Cao, Zheng, additional, Zhang, Tao, additional, Ye, Jianxi, additional, Wang, Shaochuang, additional, Feng, Fei, additional, Zhao, Li, additional, Liu, Xiaoyong, additional, Song, Liuyihan, additional, Peng, Liwei, additional, Guo, Yiqun, additional, Jiang, Xiaowei, additional, Tang, Lingbo, additional, Du, Yin, additional, Zhang, Yingya, additional, Pan, Pan, additional, and Xie, Yuan, additional

Published
2020
Full Text
View/download PDF

50. [Effect of recombinant human parathyroid hormone 1-34 on mandibular distraction osteogenesis in rabbits]

Author

Peng, Liwei and Wei, Xiaoling

Subjects
Random Allocation, Bone Regeneration, Bone Density, Parathyroid Hormone, Osteogenesis, Distraction, Animals, Humans, Mandible, Rabbits, Recombinant Proteins
Abstract

To explore the effect of recombinant human parathyroid hormone 1-34 [rhPTH(1-34)] on bone regeneration rabbit mandible during distraction osteogenesis (DO).40 Japanese white rabbit (weight 2.0-2.5 kg) were randomly divided into control group and groups. The experimental groups were divided inito 12.5, 25 and 50 µg/kg group according to the dosage of rhPTH (1-34) in each group. Each group involved 10 rabbits, and unilateral DO models were established at the right mandible of the rabbits. From the first day of distraction to the day of execution, the rabbits in the experimental groups were injected subcutaneously rhPTH (1-34) of the corresponding dose respectively, and the rabbits in the control group were injected subcutaneously 2% heat inactivated rabbit serum 1 ml respectively.. Five rabbits in each group were executed respectively at 1 week and 3 weeks after completion of distraction, and the specimens of DO were harvested. The gross observation, X-ray examination, and histological study were performed.Gross appearance: At the first week of consolidation, the dense and opaque white tissue was seen in the distraction gap of the 50 µg/kg group, and the white translucent tissue was seen in the distraction gaps of the rest groups. At the third week of consolidation, the greyish white tissue was seen in the distraction gap of the control group, while the cartilage-like tissue was seen in the buccal side of the distraction gap of the 12.5 µg/kg group, the color of new-formed tissues was close to that of normal bone tissue in the lingual side. The buccal tissue at the edge of the distraction gap of the 25 µg/kg group fitted together with the primary bone tissue in its two sides. It was difficult to distinguish the boundaries between the distraction gap and the bone tissues in its two sides in the 50 µg/kg group. X-ray findings: At the first week of consolidation, a sparse opaque image was seen in the distraction gap of the 50 µg/kg group, and a low-density image was seen in the distraction gap of the rest groups. At the third week of consolidation, a sparse bone image was seen in the control group, and the edge of the bone was not continuous. With the increase of the dose in the experimental groups, the image of the distraction gap became more and more opaque, and the image of the distraction gap in the 50 µg/kg group was close to that of the normal bone tissue. HISTOLOGICAL FINDINGS: At the first week of consolidation, few osteoblasts were present at the edge of the distraction gap of the control group. A large number of bone cells and bone trabecular were present in the distraction gap of the 12.5 µg/kg group, the network of the bone trabecula was present in the 25 µg/kg group, and a few new bones were found in the 50 µg/kg group. At the third week of consolidation, the network of the trabecular bone was present in the distraction gap of the control group, while the network of the bone trabecula was present in the 12.5 µg/kg group, a lot of bone-like tissues in the 25 µg/kg group, and near-mature bone in the 50 µg/kg group.rhPTH(1-34) can promote the formation of new bone in the distracted gap during mandibular DO in rabbits.

Published
2016

Catalog

Books, media, physical & digital resources
See catalog results