Label Recovery and Trajectory Designable Network for Transfer Fault Diagnosis of Machines With Incorrect Annotation

The success of deep transfer learning in fault diagnosis is attributed to the collection of high-quality labeled data from the source domain.However,in engineering scenarios,achieving such high-quality label annotation is difficult and expensive.The incorrect label annotation produces two negative effects:1)the complex decision boundary of diagnosis models lowers the generalization performance on the target domain,and2)the distribution of target domain samples becomes misaligned with the false-labeled samples.To overcome these negative effects,this article proposes a solution called the label recovery and trajectory designable network(LRTDN).LRTDN consists of three parts.First,a residual network with dual classifiers is to learn features from cross-domain samples.Second,an annotation check module is constructed to generate a label anomaly indicator that could modify the abnormal labels of false-labeled samples in the source domain.With the training of relabeled samples,the complexity of diagnosis model is reduced via semi-supervised learning.Third,the adaptation trajectories are designed for sample distributions across domains.This ensures that the target domain samples are only adapted with the pure-labeled samples.The LRTDN is verified by two case studies,in which the diagnosis knowledge of bearings is transferred across different working conditions as well as different yet related machines.The results show that LRTDN offers a high diagnosis accuracy even in the presence of incorrect annotation.

Improved genome annotation of Brassica oleracea highlights the importance of alternative splicing

Brassica oleracea has been developed into many important crops,including cabbage,kale,cauliflower,broccoli and so on.The genome and gene annotation of cabbage(cultivar JZS),a representative morphotype of B.oleracea,has been widely used as a common reference in biological research.Although its genome assembly has been updated twice,the current gene annotation still lacks information on untranslated regions(UTRs)and alternative splicing(AS).Here,we constructed a high-quality gene annotation(JZSv3)using a full-length transcriptome acquired by nanopore sequencing,yielding a total of 59452 genes and 75684 transcripts.Additionally,we re-analyzed the previously reported transcriptome data related to the development of different tissues and cold response using JZSv3 as a reference,and found that 3843 out of 11908 differentially expressed genes(DEGs)underwent AS during the development of different tissues and 309 out of 903 cold-related genes underwent AS in response to cold stress.Meanwhile,we also identified many AS genes,including BolLHCB5 and BolHSP70,that displayed distinct expression patterns within variant transcripts of the same gene,highlighting the importance of JZSv3 as a pivotal reference for AS analysis.Overall,JZSv3 provides a valuable resource for exploring gene function,especially for obtaining a deeper understanding of AS regulation mechanisms.

Meibomian glands segmentation in infrared images with limited annotation

●AIM:To investigate a pioneering framework for the segmentation of meibomian glands(MGs),using limited annotations to reduce the workload on ophthalmologists and enhance the efficiency of clinical diagnosis.●METHODS:Totally 203 infrared meibomian images from 138 patients with dry eye disease,accompanied by corresponding annotations,were gathered for the study.A rectified scribble-supervised gland segmentation(RSSGS)model,incorporating temporal ensemble prediction,uncertainty estimation,and a transformation equivariance constraint,was introduced to address constraints imposed by limited supervision information inherent in scribble annotations.The viability and efficacy of the proposed model were assessed based on accuracy,intersection over union(IoU),and dice coefficient.●RESULTS:Using manual labels as the gold standard,RSSGS demonstrated outcomes with an accuracy of 93.54%,a dice coefficient of 78.02%,and an IoU of 64.18%.Notably,these performance metrics exceed the current weakly supervised state-of-the-art methods by 0.76%,2.06%,and 2.69%,respectively.Furthermore,despite achieving a substantial 80%reduction in annotation costs,it only lags behind fully annotated methods by 0.72%,1.51%,and 2.04%.●CONCLUSION:An innovative automatic segmentation model is developed for MGs in infrared eyelid images,using scribble annotation for training.This model maintains an exceptionally high level of segmentation accuracy while substantially reducing training costs.It holds substantial utility for calculating clinical parameters,thereby greatly enhancing the diagnostic efficiency of ophthalmologists in evaluating meibomian gland dysfunction.

The History of the Former Han Dynasty:A Critical Translation with Annotations Translated by Homer Hasenpflug Dubs

Translation is an important medium of cultural communication.It is not a mere transfer of two languages,but the interaction of two cultures.Cultural misreading,which results from cultural discrepancy and translator’s subjectivity,truly reflects where the blockade and conflict in the cultural communication is.Cultural misreading is an objective phenomenon that exists in the entire process of translation.This paper intends to make a comprehensive analysis and discussion on The History of the Former Han Dynasty:a Critical Translation with Annotations translated by Homer Hasenpflug Dubs.As for the reasons of cultural misreading,this paper divides them into three types—language,thinking habit,traditional culture.It is to be hoped that this paper will draw more attention from the translation circle to the phenomena,and make contribution to the development of literary translation.

A Study on Annotations Under the Theory of Paratext in Translation

This paper mainly studies the basic types of annotation and the analysis of its effective functional usage,so as to pay more attention to annotation in the translation of poetry and Fu.The annotation of this study belongs to the category of paratext.Annotation is attributed to the paratext,undertakes its special function,enriches and perfects the paratext system.

Automatic Image Annotation Using Adaptive Convolutional Deep Learning Model

Every day,websites and personal archives create more and more photos.The size of these archives is immeasurable.The comfort of use of these huge digital image gatherings donates to their admiration.However,not all of these folders deliver relevant indexing information.From the outcomes,it is dif-ficult to discover data that the user can be absorbed in.Therefore,in order to determine the significance of the data,it is important to identify the contents in an informative manner.Image annotation can be one of the greatest problematic domains in multimedia research and computer vision.Hence,in this paper,Adap-tive Convolutional Deep Learning Model(ACDLM)is developed for automatic image annotation.Initially,the databases are collected from the open-source system which consists of some labelled images(for training phase)and some unlabeled images{Corel 5 K,MSRC v2}.After that,the images are sent to the pre-processing step such as colour space quantization and texture color class map.The pre-processed images are sent to the segmentation approach for efficient labelling technique using J-image segmentation(JSEG).Thefinal step is an auto-matic annotation using ACDLM which is a combination of Convolutional Neural Network(CNN)and Honey Badger Algorithm(HBA).Based on the proposed classifier,the unlabeled images are labelled.The proposed methodology is imple-mented in MATLAB and performance is evaluated by performance metrics such as accuracy,precision,recall and F1_Measure.With the assistance of the pro-posed methodology,the unlabeled images are labelled.

ASE-ERNIE:一种基于ERNIE的中文指令动作序列生成方法

中文结构灵活多样,同样的语义有多种表达方式,在中文指令过长或存在多组动作的情况下,家庭服务机器人很难正确理解并做出响应。此外,家庭环境的中文指令解析研究在数据方面仍然稀缺。现有的方法不能准确提取出指令中包含的全部动作序列。由此提出一种基于ERNIE方法的动作序列提取(ASE-ERNIE)模型来提取中文指令中的可执行动作序列,该模型通过动作类型和参数之间的关系提取指令信息,将动作序列的提取工作转化为序列标注任务。在指令中存在多组动作并且存在多个参数的情况下,依然能够提取出全部动作序列。通过实验验证了所提方法的有效性,动作序列提取任务的F 1值达到了80.37%。

BIM模型三维出图关键技术研究与应用

为解决BIM模型三维正向设计出图问题,基于MicroStation三维设计平台,通过二次开发方式系统研究成图及标注关键技术。通过提取剖切位置处模型切面、剖视范围内三维模型,实现剖面图底图绘制;通过将材料属性信息和剖面图关联,并配合预定义填充符号库,实现剖面图自动填充;通过将自定义数据结构在模型和图纸之间关联,实现图纸与模型联动更新;通过修改剖视图中前视范围内不可见边显示样式,实现二维剖视图中不可见边虚线显示;研发二、三维图面高程、坡比、桩号等快速标注、图框快速调用工具,实现快速成图。实例应用结果表明,该方法可明显提升基于BIM模型的二维图纸生成效率,适用直线型、折线型等不同剖面位置线图纸批量剖切,满足结构模型、地形模型、地质模型等同时剖切需求、不同出图比例图纸剖切要求、模型修改后图纸批量更新需求,研究方法和软件开发思路可为同类项目提供参考。

ChatGPT视域下的知识限度与积极物化:从庄周出发的思考

ChatGPT可以被看成是一个“自本自根”的世界模型。在其内部的神经网络之中,似乎存在一个道家所强调的道。然而,从《庄子·知北游》中的两段经典对话中我们获得的新启示是,ChatGPT所表达的那些可言说的知识未必是真实可靠的知识。同时,使用者与ChatGPT的对话往往会演变为一方对另一方的臣服或对抗,而日常生活中的多人对话往往可以提供一种对结论的开放性。大模型的无限知识同样存在进一步的进化空间。与ChatGPT相比,GPT-4明显知道了自己知识的限度。目前的大模型框架更多是一种普遍性框架,其对问题的回答往往是通用性的解释,而进化的方向恰恰是要增加向量数据库。知识限度的意义就在于要回到多样性本身。人类对大模型的有待使用,最终会产生无待焦虑。换言之,当外在依赖变成内在依赖的时候,人们就会产生无待焦虑。这种无待焦虑最终会演化为后人类焦虑。破除这些焦虑的关键,同样需要在庄周思想中挖掘。这种积极物化的关键是,放低身段,将自己置于一个与物平等的角度,向物学习,格物致知,最终达到物我两忘的真人之境。从庄周丧我的角度可以更加接近物我两忘,其最终可能会帮助人类进入人机和谐相处的两行世界。

领域大语言模型下的古籍词性标注应用研究

[目的/意义]大语言模型的发展为古籍文本挖掘带来了新的思路,将大语言模型与古籍数字化、智能化相结合是新时代古籍工作的必经之路。[方法/过程]文章使用《左传》词性标注语料,通过数据清洗和预处理,构建了一批高质量的词性标注指令数据,在此基础上,分别使用500、1 000、2 000、5 000条数据对大语言模型进行指令微调,并在另外1 000条数据上进行性能测试。[结果/结论]实验结果表明,“荀子”系列模型在古籍文本词性标注任务上性能优于通用领域模型,在微调数据量达到5 000时,Xunzi-Baichuan2-7B模型表现出了最优性能,F1值达到81.67%。

名实难副——从《庄子》“内七篇”看庄子的语言游戏及解构策略

解构作为一种古老的智慧并非德里达专有。天然解构大师庄子在《庄子》“内七篇”里以自由游戏的姿态切断了语言确定性的根基,并假借一群“名不副实”之人在“名”与“实”间进行自由游戏,从而揭露了“名”与“实”、能指与所指之间的矛盾和不一致。这一东方解构范例在传统庄子诠释学之外为我们提供了一个活泼且具有现代意义的庄子形象。同时,庄子对名言枷锁的解构,也使得儒道精神大放异彩,揭开了中国古代文化哲学儒道兼宗的序幕。

基于特征图网络和多种生物信息预测关键蛋白质的深度学习框架

针对生物实验识别关键蛋白质费时费力,使用计算方法预测关键蛋白质无法有效整合生物信息的问题,提出一个深度学习框架.首先利用网络拓扑结构、基因表达数据和GO(gene ontology)注释数据构建加权蛋白质相互作用网络;然后分别使用特征图网络和双向长短期记忆细胞从亚细胞定位数据、蛋白质复合物数据和基因表达数据中提取特征向量;最后将这些特征向量输入到任务学习层预测关键蛋白质.实验结果表明,相比于现有的计算方法,该方法预测性能更好.

基于自然标注的跨平台虚拟账号关联方法研究

随着大数据时代的到来,跨平台虚拟账号的关联成为网络监管领域亟待解决的问题。该文以微博、微信等用户文本数据为研究对象,通过对数据的抽样和人工标注,开展开放式社交平台中跨平台账号自然标注行为的量化分析,并由此提出基于用户自然标注的跨平台虚拟账号的关联方法。该方法针对自然标注特点,构建基于上下字词特征的虚拟账号识别的模型,并利用二分类的深度学习模型进行昵称和用户的同一认证,最终实现对跨平台虚拟账号的识别,识别准确率达到85%以上。

方以智与晦山戒显的青原之会

康熙五年(1666),晦山在青原山净居寺拜会了方以智。方以智非常兴奋,给予了极其热情的接待。晦山戒显对此感动不已,遂赋诗一首,记录下了自己的兴奋之情。方以智与晦山谈及自己的著作《药地炮庄》,晦山为之作序,给予高度评价。可以说,晦山戒显与药地愚者方以智青原之会具有多方面的文化意蕴。

花椒根腐病拮抗菌株W-1基因组测序及抑菌机理研究

【目的】分析贝莱斯芽孢杆菌(Bacillus velezensis)W-1的全基因组序列信息,探究其对花椒根腐病的生防机理,为该菌的高效开发和应用提供生物信息学基础,并为该菌开发为生物农药提供理论支持。【方法】采用三代PacBio平台测序技术对菌株W-1进行全基因组测序,并对测序结果进行基因功能注释和比较基因组学分析;采用菌丝生长速率法测定菌株W-1发酵液乙酸乙酯提取物对花椒根腐病病原菌菌丝形态、病原菌孢子悬浮液电导率和核酸含量的影响。【结果】菌株W-1基因组全长4166284 bp,GC含量为46.32%,其中编码蛋白基因4037个;GO、eggNOG和KEGG数据库中注释到的基因数分别为2935、3087和2185个;预测到菌株W-1能产生14种次级代谢产物合成基因簇,包括sur‐factin、butirosin A/butirosin B、planttazolicin、macrolactin H、bacilaene、fengycin、diffcidin、bacillibactin和bacilysin等9种已知基因簇及5种未知基因簇;菌株W-1特有的基因家族53个,特有基因414个,与模式菌株B.velezensis FZB42的亲缘关系较近。菌株W-1提取物对花椒根腐病病原菌菌丝生长具有明显的抑制作用,最小抑菌浓度(MIC)为4.50 mg/mL、最小杀菌浓度(MFC)为9.00 mg/mL。菌株W-1提取物可造成病原菌菌丝内含物外渗,在提取物浓度9.00 mg/mL处理10 h后,病原菌孢子悬浮液相对电导率比空白对照高57.63%、核酸含量比空白对照高64.91%。【结论】贝莱斯芽孢杆菌W-1能产生多种抗菌物质,其发酵液乙酸乙酯提取物可破坏花椒根腐病病原菌菌丝细胞膜的完整性,具有开发成生物农药的潜力,在花椒根腐病的绿色防控中具有良好的应用前景。

基于Transformer的司法文书命名实体识别方法

命名实体识别是自然语言处理领域的关键任务之一,是实现下游任务的基础。目前针对司法领域的相关研究相对较少,司法系统的信息化和智能化转型仍有许多问题亟需解决。相比其他领域的文本,司法文书存在专业性强、语料资源少等局限,导致现有的司法文书识别结果较低。因此,从以下3方面开展研究:首先,提出了一种多标签层级迭代的文本标注方式,可以对原始司法文书文本进行自动化标注,同时有效地提升司法文书命名实体识别任务的实体识别效果;其次,提出了一种交融式的Transformer神经网络模型,对汉字固有属性的深层特征进行了充分利用,用于对司法文书进行命名实体识别;最后,对所提出的标注方法和模型与其他神经网络模型进行了对比实验。所提出的文本标注方式可以较为准确地实现司法文书的标注任务;同时,所提出的模型在通用数据集中相对于对照模型有较大的提高,并在司法领域数据集中取得了良好的效果。

宋人注杜的三种整理本考察

宋代杜诗学极盛,有“千家注杜”之誉。宋人对于杜诗的注释,是杜诗学史上的开创和奠基之作。上海古籍出版社出版的三种宋人杜诗注释的整理本,分别代表了古籍整理的三种类型:《杜诗赵次公先后解辑校》,是在原本散佚较多的情况下辑佚而成的整理著作;《新定杜工部草堂诗笺斠证》,是在原本基础上进行了正本清源、斠证发微、订讹掘隐工作,而与原本相较,成为了一种融合较多整理者见解的新著;《新刊校定集注杜诗》,则是恪守古籍整理规范,重在校点勘误、旨在恢复原貌的精审著作。

人工智能囊胚形态评估数据集构建与质控专家共识

囊胚形态人工智能(AI)评估是AI医疗器械发展的新兴方向,也是AI在辅助生殖领域的重要应用。AI在新领域应用的起步阶段,数据集的构建与质控对产品质量有重要影响。目前,囊胚形态学AI评估在数据采集、标注、质控等方面尚未形成统一的规范。在参考AI医疗器械、辅助生殖医疗器械现有国家行业标准的基础上,本文以囊胚形态AI评估数据集为主题,对数据集构建与质控要求进行了探讨,对数据集质量特性进行了解析,旨在指导数据集制造责任方加强数据集全生命周期管理,更好地为产品研发、测试、临床试验等环节提供质量保障,助力产业发展。

《庄子·齐物论》“彼是莫得其偶”的涵义及现代价值

《齐物论》是《庄子》中极其重要的篇章,从篇名到具体内容无不蕴含着丰富的哲学思想,其中“彼是莫得其偶”是一个很重要的命题,是实现“吾丧我”境界的重要环节,理解这一命题的关键在于对“彼是”的理解。“彼是”既不是“彼此”,也不是“是非”,而是“彼此”“是非”,“彼此”即“是非”。“彼是莫得其偶”有去自我中心主义和去人类中心主义两方面价值。

基于自监督的主动标签清洗

主动标签清洗利用主动学习来进行标签噪声处理,以降低人工标注成本。现有的主动标签清洗方法仍然存在人工额外标注成本较高的问题,即挑选出的可疑样本中正确样本所占比例较高。为了缓解这一问题,提出了一种基于核心集的自监督主动标签清洗方法。首先利用自监督任务进行表征学习,随后将数据映射到特征空间中,并利用贪婪的K-Center集合覆盖方法挑选出可疑样本,最后根据不确定性筛选出标签噪声样本进行重标注。并同时考虑到了样本的代表性与不确定性,能够有效降低可疑样本中正确样本的比例。在含有不同比例标签噪声的公开数据集上的实验结果表明,在各迭代轮次中明显地降低了人工额外标注成本,同时也在一定程度上缓解了冷启动问题。此外,还通过消融实验证明了方法中自监督核心集采样模块和不确定性预测模块的有效性。