一种基于类不平衡学习的情感分析方法

针对网络评论中普遍存在的负面评论较少而影响力却较大的类不平衡问题, 提出一种基于类不平衡学习的情感分析方法. 该方法利用深度学习训练过程中的概率输出, 以计算样例的信息熵作为影响因子构建交叉信息熵损失函数. 在IMDB公开数据集上进行实验验证的结果表明, 基于集成信息熵损失函数的双向长短期记忆网络能处理类不平衡问题; 对数据的统计分析结果表明, 该策略能提升基于双向长短期记忆网络的评论情感极性分类性能. 针对AUC(area under curve)指标, 使用集成信息熵损失函数的双向长短期记忆网络模型比未考虑类不平衡的深度学习模型在中位数上最多提升15.3%.     

巨灾风险大数据处理应急分类、分解、分拣算法与应用

本文主要研究巨灾风险大数据处理的应急分类、分解、分拣算法,给出了相应的算法原理和可操作的步骤.首先根据巨灾风险大数据灾害规模巨大的特征,提出了一种用来解决巨灾风险大数据中一级事件的应急分类与二级事件及以下更低级事件的应急分解算法,并以特大地震灾害作为实例进行了算法应用.接着定义了事故灾难度,用来对巨灾风险大数据处理过程中,对各种级别的事故灾难后果进行不同的数字标识.然后提出一种用来解决巨灾风险中大数据快速处理的应急分拣算法,并在汶川地震中大规模灾害的应急救援计划中进行应用.经过采用这样的应急分拣原理,就可以在面对巨灾风险大数据的复杂、繁多和零乱的重灾事件状态下,使整个应急救援方案优化,并能够有条不紊地进行救援.     

基于新型空间注意力机制和迁移学习的垃圾图像分类算法

随着我国各级政府大力推动垃圾强制分类,分类回收各环节中实现标准化、自动化的垃圾分类识别需要适合云端部署的高准确率、低延时要求的细粒度图像分类模型.本文发挥深度迁移学习的优势建立了一套端到端的迁移学习网络架构GANet (garbage neural network);针对垃圾分类中类别易混淆、背景干扰等挑战,提出一种新型的像素级空间注意力机制PSATT (pixel-level spatial attention).为克服类别多和样本不平衡挑战,提出使用标签平滑正则化损失函数;为改善收敛速度以及模型稳定性与泛化性,提出了阶梯形OneCycle学习率控制方法,并给出了结合Rectified Adam (RAdam)优化方法和权重平滑处理技术的组合使用策略.实验使用了"华为云人工智能大赛.垃圾分类挑战杯"提供的按照深圳市垃圾分类标准标注的训练数据,验证了GANet在垃圾分类问题中的显著效果,获得了全国二等奖(第2名);同时,提出的PSATT机制优于对比方法,且在不同主干网络架构上均得到了提升,具有良好的通用性.本文提出的GANet架构、PSATT机制和训练策略不仅具有重要的工程参考价值,也具有较好的学术价值.     

基于三维点云的岩体结构面自动分类与参数计算

结构面间距是岩体稳定性和力学特性分析中的一个重要参数，在岩石力学、采矿工程、边坡监测等领域的数值计算中广泛应用.本文以岩体边坡露头为研究对象，基于非接触测量获得的三维点云数据，提出一种基于密度聚类的结构面细化分类方法；在结构面粗略分组提取的基础上，通过投影变换、散乱点拟合等算法，求得结构面间距和岩体体积节理数.设计开发了结构面细化分类及间距等参数计算与分析原型系统，实际案例分析表明，本文方法可有效实现结构面的自动细化分类，并能够计算出间距等相关参数，可为岩体质量分级和岩体稳定性分析等提供方法支撑.     

工艺规划系统中动态模式分类的研究

提出了工艺规划系统中动态模式分类（pattern classification）的概念，介绍了模式分类判定树和动态模式分类的基本原理，并对动态模式分类的聚类质量标准进行了研究。动态模式分类是基于实例归纳（case based reasoning,CBR)的自动分类过程，具有自学习和自完善的功能。     

Solving Non-Uniqueness in Agglomerative Hierarchical Clustering Using Multidendrograms

In agglomerative hierarchical clustering, pair-group methods suffer from a problem of non-uniqueness when two or more distances between different clusters coincide during the amalgamation process. The traditional approach for solving this drawback has been to take any arbitrary criterion in order to break ties between distances, which results in different hierarchical classifications depending on the criterion followed. In this article we propose a variable-group algorithm that consists in grouping more than two clusters at the same time when ties occur. We give a tree representation for the results of the algorithm, which we call a multidendrogram, as well as a generalization of the Lance andWilliams’ formula which enables the implementation of the algorithm in a recursive way. The authors thank A. Arenas for discussion and helpful comments. This work was partially supported by DGES of the Spanish Government Project No. FIS2006–13321–C02–02 and by a grant of Universitat Rovira i Virgili.     

Developmental modes of the Neoproterozoic-Lower Paleozoic marine hydrocarbon source rocks in China

Based on the researches on rock type, the mode of occurrence, diagnostic minerals and creatures, the sedimentary geochemistry and organic facies of Chinese marine source rocks from wells and outcrops, and on the research findings of developmental modes of foreign marine source rock, the authors consider that it is impossible to objectively make clear the formation mechanism of hydrocarbon source rock with high organic matter abundance by either single mode of preservation or high organic matter productivity. According to the Chinese geological features, the formation mechanism of the Neoproterozoic-Lower Paleozoic marine source rock is generalized into four modes, namely, thermal water activity-upwelling flow-anoxic; carbonate gentle slope-upwelling flow; xerothermic climate-brine euxinic milieu; and humid climate-retained euxinic milieu; as the Lower Cambrian undercompensation basin organic facies （the Tarim Basin, South China and southwestern margin of North China）, carbonate gentle lime mud bound organic facies （the Upper Ordovician in Tazhong region of the Tarim Basin and the Lower Silurian in the Upper Yangtze Platform）, the Middle Cambrian evaporation laggon organic facies （the Tarim Basin and the Upper Yangtze Platform）, enclosed undercompensation terrigenous bay organic facies （the Middle-Upper Ordovician in the west of the Tarim Basin, the Lower Silurian Long- maxi Formation in the Upper Yangtze platform）. Chinese marine sedimentations with lower organic matter abundance are generalized into two modes of consumption-dilution mode of open epicontinental sea and depletion-dilution mode of supercompensation basin.     

An efficient algorithm to generate candidates in discovering frequent episodes

0 IntroductionData mining is widely used in many research fieldssuch as decision supporting systems[1], bio-informationanalysis[2]and knowledge engineering[3-5]. Most data col-lected from scientific experiments or telecommunicationnetworks have inherent sequential nature inthemand canthus be abstractly viewed as a sequence of events . Onebasic problemin miningsuchevent sequencesis discoveryof recurrent combinations of events , which are calledepisodes. Once frequent episodes are discovered,rul…     

基于DELPHI5的高校选修课管理系统的设计与实现

主要介绍高校选修课管理系统的设计思想，并着重分析和讨论了后台数据库的设计方法和原理，最后讨论了运用Delphi5开发该系统的整个过程。