基于BERT-SELFATT-CNN模型的垃圾邮件分类方法

针对传统垃圾邮件分类方法中使用静态词向量不能解决一词多义、长序列信息特征提取不足等问题,提出了一种基于BERT-SELFATT-CNN模型的垃圾邮件分类方法。使用动态文本表示方法BERT对邮件内容进行预训练,并生成带有上下语义信息的词向量,经过能够并行计算的自注意力机制层计算词与词之间的相似度去挖掘句子长距离信息,将生成的隐藏层向量输入到CNN网络提取向量局部特征。在中文垃圾邮件数据集上与已有模型进行对比实验,结果表明该模型在精确度、召回率和F1值上均有提高,模型训练速度也得到提升。

基于并行CNN-Self attention&LSTM的锂电池RUL间接预测

针对锂离子电池难以在线测量直接健康因子(容量、内阻)进行剩余使用寿命(RUL)预测的现状,提出基于并行CNN-Self attention与LSTM组合的锂电池RUL间接预测方法。首先基于锂离子电池放电循环数据,利用灰度关联分析筛选出能够高度体现电池容量衰减的间接健康因子;然后分别对间接健康因子使用卷积神经网络(CNN)进行特征提取和自注意力机制(Self attention)调整特征权重得到对应的特征矩阵;最后融合特征矩阵,并使用长短期记忆(LSTM)网络进行锂电池RUL预测。使用NASA锂离子电池数据集进行实验,结果表明:与传统的CNN、LSTM和CNN-LSTM模型相比,并行CNN-Self attention&LSTM模型所得到的容量衰减曲线更接近实际容量衰减曲线,误差波动范围更小,在RMSE、MAE和R2系数3种评价指标上表现更优。

认知视角下英语"-self"和汉语"自己"对比研究

在认知语言学框架内的原型范畴理论指导下,从词类属性、泛指功能、语篇分布特征等角度对英汉反身代词进行对比研究。研究表明:英汉反身代词均为即指语和离指语这一连续统一体内的成员,但二者在其中的地位并不相同;在功能方面,虽然英汉反身代词都具有泛指功能,但前者的使用是有条件的,后者的使用是无限制的;在语篇分布特征上,汉语反身代词在语篇中的使用频率要高于英语中的反身代词。

优选论理论下英语“X-self”和汉语“自己”的对比研究

英语"X-self"和汉语"自己"都可以出现在宾语语域,但二者在充当宾语时所表现出来的约束性不同。在优选论指导下,制定了二者在句型"NP1+V1P1(PP1)+NP2+V2P2+R"中的约束等级,并考察了影响二者约束性的各种因素之间的关系。结果表明:英语"X-self"的约束等级和汉语"自己"的约束等级既含有不同的原则也含有相同的原则;正是由于不同的原则,造成了在英语和汉语中句法结构完全相同的句子而二者约束性的不同。

基于多特征LSTM-Self-Attention文本情感分类

针对自然语言处理情感分析领域中情感分类的问题,提出了一种基于多特征LSTM-Self-Attention的文本情感分类方法。方法以词向量和词性向量为输入,利用LSTM网络模型提取文本的序列特征,并通过在模型中引入自注意力机制(self-attention),从序列特征中提取出句子的语法和语义特征,减少了任务的复杂度。上述方法避免了传统循环神经网络存在的梯度消失和梯度爆炸的问题,极大缩短单词长距离依赖特征之间的距离,提高了分类效果。最后使用中文电影评论数据集进行实验验证,结果表明该方法特征提取能力更强,使得情感分类的准确率提升了1.74%。

THE RANDOM SHIFT SET AND RANDOM SUB-SELF-SIMILAR SET

First of all the authors introduce the concepts of random sub-self-similar set and random shift set and then construct the random sub-self-similar set by a random shift set and a collection of statistical contraction operators.

THE DIMENSION FOR RANDOM SUB-SELF-SIMILAR SET

In this article, the Hausdorff dimension and exact Hausdorff measure function of any random sub-self-similar set are obtained under some reasonable conditions. Several examples are given at the end.

MultiHead-SelfAttentionBiLSTM网络应用于主观题评分的研究

为了降低主观题评分对人工阅卷的依赖,提升主观题自动阅卷的效率和准确性,使评分结果更客观,设计了MultiHead-SelfAttention BiLSTM Automatic Grading模型。该模型是将深度学习与多头自注意力机制结合,采用BiLSTM模型对进行过人工判分的主观题进行文本语义特征学习;利用多头机制模拟多人主观判分从多个角度进行关系抽取,降低阅卷偏差;关系抽取利用自注意力机制提取出符合得分要求的关键特征组合。实验结果表明,MultiHead-SelfAttention BiLSTM Automatic Grading模型与常规方法比较,对主观题自动评分的准确率均有不同程度的提高,精确率、召回率、F1值均达到了较好的效果。对减少人工阅卷成本、推进中文自动化评判的进一步发展起到了积极作用。

Research on spraying and fusing of non-self-fluxing copper powdered alloy on cast iron surface

A new type additive was added to the non-self-fluxing copper powdered alloy, and the powder showed satisfactory performance of spraying and fusing, self-protecting, and self-fluxing in the oxygen-acetylene flame spraying and fusing process. The additive was melted and could absorb copper oxide when spraying, then it created a layer of film to cover the copper liquid, which protected the copper liquid from oxidizing efficiently and promoted it wetting on iron surface. Thus it lead to real diffusion between Cu and Fe, which resolved the diffwulty of combining Cu with Fe and reduced the limitation of the powder, and it promoted the usage value of general powder. Through analysis of microstructures, it was found that the fastness and compactness of the combining interface was excellent as well as the diffusing of transition area, and the hardness was suitable, which meant that the performance could meet the needs of high quality valves completely.

NONEXISTENCE OF PSEUDO-SELF-SIMILAR SOLUTIONS TO INCOMPRESSIBLE EULER EQUATIONS

In this paper we study a generalization of self-similar solutions. We show that just as for the solutions to the Navier-Stokes equations these supposedly singular solution reduce to the zero solution.

CRITERIA OF STRONG TRANSIENCE FOR OPERATOR-SELF-SIMILAR MARKOV PROCESSES

Yamamuro in [1] defines strong and weak transience of Markov processes; gives a criterion for strong transience of Feller processes; and further, discusses strong and weak transience of Ornstein-Uhlenbeck type processes. In this article, the authors weaken the Feller property of the result in [1] to weak Feller property and discuss the strong transience of operator-self-similar Markov processes.

基于阴性选择原则的Non-self探测器生成算法

基于免疫系统异己检测原理,深入进行了计算机免疫系统探测器生成算法的研究.首先,简要介绍了阴性选择算法,总结了相关的探测器生成算法;然后,基于阴性选择原则提出了两种探测器生成算法,即位变异算法(BMGDGA)和余数生长算法(AGDGA).文中对两种算法在多种不同的数据集上进行了全面的验证和实验,并与穷尽式探测器生成算法进行了全面系统的比较.结果表明,两种探测器生成算法在综合性能上均优于穷尽式探测器生成算法.

A New Compact Omni-Directional Quasi-Self-Complementary Antenna for Wireless Communication Systems

This article describes a new miniaturized omni-directional antenna with quasi-self-complementary structure for wireless communication applications. A novel ground structure composed of five rectangular plates is proposed to enhance the impedance bandwidth and reduce the antenna size. The proposed antenna is comprised of two patches surrounded by the ground structure. Two metal patches of the antenna are located on two opposite sides of the dielectric substrate. The feed patch is used to excite the radiation patch. This unique design is realized by properly choosing the suitable feed patch shape, selecting similar slot shape on the radiation patch, and tuning their dimensions. The proposed antenna with an extremely small size of 6 mm × 9 mm has an operating impedance bandwidth ranging from 4.5 to 6.1 GHz for S11 < -10 dB, which also covers the two IEEE 802.11a wireless local area network bands (5.15 - 5.35 GHz and 5.725 - 5.825 GHz). In addition to be very small in size, the antenna exhibits omni-directional radiation patterns in the entire operating bandwidth and low cross polarization. The distortionless time domain performance of the antenna is confirmed by investigation of the phase response and group delay. The obtained results in both frequency and time domain show that the proposed antenna is suitable for use in wireless communication systems.

惠特曼诗歌“One's-Self I Sing”译法之我见

通过对惠特曼诗的两种译文的比较和评论,希望能够更加充分地体现原诗的"自由与民主"主题和"有节奏、无韵律"的风格,使用更简短、准确的措辞,尾韵和排比的方法,使新的译文更加准确、生动、流畅、自然。

Cardiac-self efficacy predicts adverse outcomes in coronary artery disease (CAD) patients

Background: Little is known about the predictive role of Cardiac Self Efficacy (CSE) in the ensuing months following a coronary event. We sought to determine whether CSE predicts adverse events in the months following discharge in patients with Coronary Artery Disease (CAD). Design: Data from a prospective study of 193 patients recently hospitalized for CAD. Methods: Data were collected via participant self-report and medical records at 3-month post-discharge (baseline;T1), 6-month post-discharge (T2) and 9-month post-discharge (T3). CSE was measured using the Cardiac Self Efficacy Scale. Multi-variate regression modeling was applied to explore the association between baseline CSE scores and cardiac-related hospital admissions and functional cardiac status at T2 and T3. Other outcomes included any hospital admissions, self-reported mental and physical health at follow up. Results: Higher CSE scores at baseline significantly predicted better cardiac functioning and self-rated mental and physical health at both T2 and T3 (with one exception);this was consistent across all five models. While baseline CSE did not predict cardiac or other hospital admission at T2, CSE was a significant predictor of both outcomes at T3;higher CSE scores resulted in reductions in likelihood of hospital admissions. After adjustment for psychosocial variables however, neither association remained. Baseline depression explained the association between baseline CSE and any cardiac admissions, as well as baseline CSE and any hospital admissions at T3 follow up. Conclusions: While CSE can predict key outcomes following a CAD event, much of the association can be explained by the presence of depression.

An Examination of the Validity of the Differentiation-of-Self Hypothesis and Its Mechanism Related to Negative Life Event Aversion among Japanese University Students

This study examined the validity of the differentiation-of-self hypothesis for Japanese university students, and its mechanism in relation to negative life event aversion. Although the differentiation-of-self hypothesis is accepted as one of the basic hypotheses in the clinical field of family therapy in Japan, empirical studies on the subject are scarce. Further, there are few studies, internationally, on its mechanism. A questionnaire survey was conducted with university students (n = 250) to measure their differentiation of self, maladaptive states, and negative life event aversion. Compared to the better-differentiated group, the less-differentiated group had higher scores both for maladaptive states—somatic symptoms (t(223) = -2.70, p = 0.008), social dysfunction (t(211.00) = -3.30, p = 0.001), and severe depression (t(208.78) = -4.08, p < 0.001)—and negative life event aversion—interpersonal domain (t(223) = -3.96, p < 0.001) and achievement domain (t(223) = -3.83, p < 0.001). The results supported the validity of the differentiation-of-self hypothesis for Japanese university students, and that cognitive tendencies toward negative life events may be part of the mechanism for less-differentiated individuals’ subsequent vulnerability. This suggests the effectiveness of cognitive interventions.

ConvNeXt网络及Stacked BiLSTM-Self-Attention在轴承剩余寿命预测中的应用

在滚动轴承剩余使用寿命预测方面,采用传统方法时存在鲁棒性差、精度低等各种问题。近些年来深度学习的发展为解决这些问题提供了新的思路。为了进一步提高对轴承寿命的预测精度,提出了一种基于ConvNeXt网络、堆叠双向长短时记忆网络(SBiLSTM)和自注意力机制(Self-Attention)的滚动轴承寿命预测方法。首先,采用连续小波变换(CWT)构造了振动信号的时频图,以更好地捕捉信号的时域和频域特征;然后,将得到的时频图输入到构建的ConvNeXt网络中,通过卷积、池化和层归一化等操作,对时频图的关键特征进行了提取;最后,将提取后的特征输入到SBiLSTM-Self-Attention模块中,进一步提取了时序信息和特征权重分配数据,利用PHM2012挑战数据集进行了验证,通过实验分析了该方法的均方根误差(RMSE)和平均绝对误差(MAE)。研究结果表明:相较于现有技术方法,该方法的平均RMSE为0.031;与其他三种方法,即卷积神经网络(CNN)、深度残差双向门控循环单元(DRN-BiGRU)和深度卷积自注意力双向门控循环单元(DCNN-Self-Attention-BiGRU)相比,其平均RMSE值分别下降了79%、74%和55%,MAE值分别下降了78%、73%和53%,说明该方法在滚动轴承剩余寿命预测中有较好的性能。

大学生自杀行为影响因素的分类决策树分析

为了识别大学生自杀行为的风险和保护因素。对在校大学生实施感恩问卷、自我控制双系统量表、自杀问卷及人口学信息调查,获得有效数据11288份,进行分类决策树分析。结果显示:大学生近一年内自杀意念发生率17.1%,近半年内自杀行为发生率3.8%。分类决策树分析发现,自杀意念为根节点,感恩、冲动、控制、家庭阶层、是否独生均为重要的子节点(AUC=0.959)。结论:自杀意念是大学生自杀行为最重要的近端风险因素,冲动为中端风险因素,感恩特质和适度控制分别为远端保护因素;低社会家庭阶层、中高阶层家庭的非独生子女大学生由自杀意念转变自杀行为的风险较高。

自我威胁启动下不同类型高自尊大学生对攻击性词语的注意偏向

目的:考察自我威胁启动下,不同类型高自尊大学生对攻击性词语的注意偏向特点。方法:选取650名大学生,通过外显自尊量表(SES)选出高自尊被试。然后,通过内隐自尊测验(IAT)筛选出不同类型高自尊(脆弱型、安全型)组各43人。每组分别参加不同难易程度的瑞文标准推理测验(SPM)完成自我威胁启动,然后完成空间线索实验。当线索无效时,根据两组对攻击性词语的反应时与中性词语的反应时差值(RTI)的变异程度,获得注意偏向特点。结果:在高自我威胁启动下,脆弱型高自尊组RTI值高于安全型高自尊组(P<0.01);在低自我威胁启动下,脆弱型高自尊组和安全型高自尊组的RTI值差异无统计学意义(P>0.05)。结论:相较于安全型高自尊,脆弱型高自尊更易受自我威胁的影响,其在高自我威胁启动下,注意偏向表现为注意固着。

创造力与大学生生命意义感:积极情绪和创造性自我效能感的作用

为考察创造力与大学生生命意义感之间的关系,以及积极情绪和创造性自我效能感在其中的作用,本研究采用Runco创新观念行为量表、积极情绪量表、创造性自我效能感问卷和生命意义问卷对665名大学生进行了测查。结果显示:(1)创造力、积极情绪、创造性自我效能感和生命意义感之间存在两两正相关关系;(2)积极情绪在创造力与生命意义感之间起部分中介作用;(3)创造性自我效能感调节了创造力影响生命意义感的直接路径,随着创造性自我效能感的提升,创造力对生命意义感的预测作用减弱。本研究在理论上揭示了创造力对生命意义感的作用及其机制,对大学生生命意义教育具有一定实践参考价值。