检索结果-南通市图书馆

Face Attributes Extraction with Muti-Task Learning and Knowledge Distillation

在线全文

学校读者我要写书评

暂无评论

Face Attributes Extraction with Muti-Task Learning and Knowl...

第40届中国控制会议

作者： Gaowu Xu Jun Li Yang Du Wenjie Shu zixuan lan Department of Automation University of Science and Technology of China Institute of Advanced Technology University of Science and Technology of China

The image of a human face contains a lot of key information, and extracting these information can effectively help us understand the facial feature information. In this paper, we propose a multi-task classification network for human faces, which is used to detect the key information of human faces. We use ResNet18 as the basic network and add an attention mechanism to each residual block. Our model includes seven binary classification branches and a regression prediction branch, these branch networks share the same parameters of the backbone network. Finally, we use the method of knowledge distillation to compress the network and reduce the amount of parameters. Since there is no suitable public dataset for our task, we built a dataset containing 76751 images ourselves. we have obtained excellent results in our face dataset. The accuracy of five binary classifications is above 90%, the accuracy of the remaining two binary classifications is above 85%, and the average error of continuous attribute prediction is 3.34.

关键词： Muti-Task Learning image classification attention mechanism knowledge distillation

Text Classification under Attention Mechanism Based on Label Embedding

在线全文

cnki会议

学校读者我要写书评

暂无评论

Text Classification under Attention Mechanism Based on Label...

第40届中国控制会议

作者： Guorui Yuan Jun Li zixuan lan Wenjie Shu Gaowu Xu School of Information Science and Technology University of Science and Technology of China Institute of Advanced Technology University of Science and Technology of China

Text classification is one of key tasks for representing the semantic information of documents. There have been a lot of studies that applied Long Short-Term Memory（LSTM） to this task. This paper develops a novel model to improve the traditional LSTM for text classification, which is based on label embedding and attention mechanism. We use Convolutional Neural Network（CNN） to learn the compatibility between words and labels, which is used as attention value to improve the sequence output generated by Bi-LSTM to ensure that the weight of related words is higher than that of unrelated *** results on five text datasets demonstrate that the proposed architecture is very competitive compared with the stateof-the-art models.

关键词： Text Classification Attention Mechanism Label Embedding Bi-LSTM

Water Level Prediction Based on Improved Spectral Residual Preprocessing and Convolution Neural Network

在线全文

cnki会议

学校读者我要写书评

暂无评论

Water Level Prediction Based on Improved Spectral Residual P...

第40届中国控制会议

作者： zixuan lan Jun Li Wenjie Shu Gaowu Xu Jun Nie Shengqiang Liu Department of Automation University of Science and Technology of China Institute of Advanced Technology University of Science and Technology of China

Time series prediction is one of the important tasks of data mining,which provides key guidance for a lot of work,such as financial stock prediction,water level prediction,and so *** preliminary work of time series prediction includes data anomaly detection,missing time series repair,and other *** at all kinds of water level and discharge time series data around Chaohu lake,this paper proposes a convolution neural network prediction model combined with a significance detection *** preprocessing module of the model makes use of the improved significance detection algorithm of computer vision and a principle of electric field *** the water level and discharge of each station around Chaohu lake from June to August 2020 as convolution neural network（CNN）input and the water level of Zhongmiao station in Chaohu lake in the next 5 days as prediction output,an excellent prediction result with MAE of 0.056m is *** the same time,it is proved that the computer vision algorithm is feasible in dealing with one-dimensional time series.

关键词： Lake water level prediction spectrum residual convolution neural network Chaohu lake

Missing Value Imputation and Prediction of River Water Quality Based on GRU–Autoencoder with Input-decay

在线全文

cnki会议

学校读者我要写书评

暂无评论

Missing Value Imputation and Prediction of River Water Quali...

第40届中国控制会议

作者： Wenjie Shu Jun Li zixuan lan Gaowu Xu Jun Nie Shengqiang Liu Department of Automation University of Science and Technology of China Institute of Advanced Technology University of Science and Technology of China

With the development of human beings, the pollution of freshwater resources occurs very frequently, which directly leads to the reduction of available water sources and the increase of maintenance costs. In order to manage and protect water resources better, a large number of sensors are used to monitor water quality. However, the problem of missing data has a negative impact on our data analysis work. In this paper, we propose a new missing value imputation model which based on GRU autoencoder with input-decay（GRUD-AE）. The prediction module is added to the output of the autoencoder to realize the short-term prediction. The model is trained with datasets which contain missing value. The input-decay module is applied to learning the time correlation of sequences which could help improve imputation accuracy. Experiments show that our model performs better than the traditional missing value imputation and prediction methods.

关键词： Missing value imputation water quality prediction GRU deep learning

超薄氧化硅与氮化钛复合层对光生电子导出的作用

在线全文

cnki会议

学校读者我要写书评

暂无评论

光学学报 2022年第13期42卷 223-227页

作者：兰自轩王艺琳赵磊马忠权上海大学物理系上海大学-索朗光伏材料与器件R&D联合实验室上海200444

采用湿法化学氧化和直流磁控溅射制备了超薄氧化硅(a-SiO)和高电子浓度的氮化钛(TiN)复合薄膜,研究其对半导体-准绝缘体-半导体(SQIS)光伏器件中晶硅表面的钝化作用。实验结果表明:相对于铝背场电极,该光伏器件的光电转换效率相对提高了... 详细信息

采用湿法化学氧化和直流磁控溅射制备了超薄氧化硅(a-SiO)和高电子浓度的氮化钛(TiN)复合薄膜,研究其对半导体-准绝缘体-半导体(SQIS)光伏器件中晶硅表面的钝化作用。实验结果表明:相对于铝背场电极,该光伏器件的光电转换效率相对提高了20.72%。结合少子寿命测量和AFORS-HET模拟软件分析,揭示了开路电压增加的原因。aSiO/TiN与n-Si有2.28 eV的价带偏移,可以减缓空穴在背部界面的复合,使复合速率降为原先的1/10,从而有效增加开路电压,达到提高异质结器件光电转换效率的目的,为SQIS光伏器件提供了一种工艺简单、制备成本低的背部钝化接触复合材料。

关键词：太阳能电池钝化层复合速率异质结空穴阻挡超薄氧化硅/氮化钛

量子绝热捷径技术在三波导耦合器设计中的应用

在线全文

学校读者我要写书评

暂无评论

上海大学学报（自然科学版） 2020年第1期26卷 69-75页

作者：韩知勇施解龙兰自轩陈玺上海大学理学院上海200444

通过将量子无摩擦动力学与绝热消除结合实现了量子绝热捷径技术在光波导耦合器中的设计.为了使该方案更实际可行,引入了幺正变换,以便对波导的宽度以及相邻波导之间的间距进行设计,并利用光束传播法对三波导耦合器进行模拟验证.模拟结... 详细信息

通过将量子无摩擦动力学与绝热消除结合实现了量子绝热捷径技术在光波导耦合器中的设计.为了使该方案更实际可行,引入了幺正变换,以便对波导的宽度以及相邻波导之间的间距进行设计,并利用光束传播法对三波导耦合器进行模拟验证.模拟结果表明,相比于绝热耦合,利用绝热捷径设计的耦合器可以在传输长度缩短4倍的情况下实现能量的有效耦合.该方案可适用于量子信息处理和集成光学领域.

关键词：波导耦合器绝热消除量子绝热捷径

交通强国建设视域下公路交通数字孪生体系架构、关键技术与实践案例

在线全文

学校读者我要写书评

暂无评论

交通运输研究 2023年第4期9卷 104-124页

作者：伍朝辉徐建达符志强蓝梓轩吕子一交通运输部科学研究院北京100029 浙江交投高速公路运营管理有限公司浙江杭州310022

交通强国建设对公路数字化转型升级提出了更高要求。数字孪生连通物理世界与数字世界,为公路交通的数字化转型落地实施提供新的路径。针对目前公路交通数字孪生应用中存在的缺少基础理论研究、应用价值分析不足、体系架构与技术体系亟... 详细信息

交通强国建设对公路数字化转型升级提出了更高要求。数字孪生连通物理世界与数字世界,为公路交通的数字化转型落地实施提供新的路径。针对目前公路交通数字孪生应用中存在的缺少基础理论研究、应用价值分析不足、体系架构与技术体系亟待深化、理论实践有待持续迭代等问题,围绕公路交通数字孪生基础理论与实践案例展开研究。在明确公路交通数字孪生内涵、外延、特征和定义等理论的基础上,分析公路典型应用场景、需求与价值,梳理适用的数字孪生平台架构和关键技术体系,对比分析典型应用案例,探讨数字孪生规模化应用面临的挑战与瓶颈,给出公路交通数字孪生可持续发展的建议。研究结果表明:公路交通数字孪生的实践应突破“物联感知”与“可视呈现”阶段,聚焦“降本增效提质”本质目标,提升“双向互动”和“模拟择优”能力,支持“人民满意、保障有力、世界前列”的交通强国建设。

关键词：交通强国建设数字化转型数字孪生虚实映射模拟择优降本增效

在线全文

学校读者我要写书评

暂无评论

中药配方颗粒质量控制的研究进展

沈阳药科大学学报 2024年第7期41卷 847-854页

作者：韩子璇陶晓倩程岚张纯刚辽宁中医药大学药学院辽宁大连116620 长治医学院药学系山西长治046000 祈蒙股份有限公司内蒙古赤峰024330

随着中药配方颗粒试点工作宣告结束、部分品种中药配方颗粒的国家标准正式发布,中药配方颗粒正式进入“标准化”的新时代。在中药配方颗粒快速发展的过程中,如何有效对其进行质量控制一直是众多学者研究的重点。文章以质量标准的现状为... 详细信息

随着中药配方颗粒试点工作宣告结束、部分品种中药配方颗粒的国家标准正式发布,中药配方颗粒正式进入“标准化”的新时代。在中药配方颗粒快速发展的过程中,如何有效对其进行质量控制一直是众多学者研究的重点。文章以质量标准的现状为切入点,分析质量标准难以建立的主要问题,并通过文献查阅总结目前中药配方颗粒的质量控制模式及未来发展的新思路,为推动中药配方颗粒的国际化、标准化发展提供参考。

关键词：中药配方颗粒质量标准质量控制多指标成分生物评价

在线全文

学校读者我要写书评

暂无评论

金昭胶囊对慢性酒精性肝损伤小鼠的保护作用

世界中医药 2024年第12期19卷 1772-1778,1786页

作者：贾岚高晶李丽邵紫萱扈觐玺刘佳琪冯广旭王景霞北京中医药大学北京100029

目的:基于Kelch样环氧氯丙烷相关蛋白1(Keap1)/核因子E2相关因子(Nrf2)/核因子κB通路研究金昭胶囊对慢性酒精性肝损伤小鼠的保护作用。方法:小鼠按体质量随机区组法分为空白组,模型组,阳性药组,金昭胶囊低、中、高剂量组,每组10只。除... 详细信息

目的:基于Kelch样环氧氯丙烷相关蛋白1(Keap1)/核因子E2相关因子(Nrf2)/核因子κB通路研究金昭胶囊对慢性酒精性肝损伤小鼠的保护作用。方法:小鼠按体质量随机区组法分为空白组,模型组,阳性药组,金昭胶囊低、中、高剂量组,每组10只。除空白组外,其余各组小鼠每日灌胃50%乙醇10 mL/kg。各组持续给药4周后,观察小鼠状态,测定血清转氨酶活性、肝组织乙醇及脂肪代谢相关酶活性、抗氧化酶活性及Keap1/Nrf2/核因子κB通路相关指标并取肝脏进行苏木精-伊红(HE)病理学检查。结果:与模型组比较,金昭胶囊各剂量组小鼠血清转氨酶活性、肝组织乙醇及脂肪代谢相关酶活性、抗氧化酶水平、炎症介质中白细胞介素-10(IL-10)水平、Nrf2 mRNA水平及Nrf2、血红素加氧酶1(HO-1)表达水平均显著升高(P<0.05、P<0.01);肝组织炎症介质中肿瘤坏死因子-α(TNF-α)、白细胞介素-6(IL-6)水平、Keap1mRNA、核因子κB亚基p65亲和肽(NF-κB p65)mRNA水平及Keap1、NF-κB p65和磷酸化核因子κB抑制蛋白α(pIκ-Bα)表达均显著降低(P<0.05、P<0.01)。结论:金昭胶囊具有保护慢性酒精性肝损伤的功效,其作用机制可能与增强肝组织中乙醇代谢酶、抗氧化酶活性,调节炎症介质水平和Keap1/Nrf2/核因子κB通路各指标mRNA及蛋白表达水平相关。

关键词：金昭胶囊酒精性肝损伤乙醇代谢脂肪代谢炎症介质氧化应激 Kelch样环氧氯丙烷相关蛋白1/核因子E2相关因子/核因子κB 小鼠

同方期刊数据库