神经网络在复杂系统建模中的应用

 分析了神经网络模型的特点、复杂系统的特征与建模方法论要求,表明神经网络应用于复杂系统建模可行且适用。探讨了当前神经网络在复杂系统建模中应用的文本语言处理、语音识别、图像处理、故障诊断等,针对所存在的不能很好地描述系统结构与演化、不能同时处理多个任务、不能很好解决非重复性问题等主要不足,提出了解决这些问题的初步思路。     

Induction of cranial and posterior trunk neural crest by exogenous retinoic acid in zebrafish

Retinoic acid (RA) plays an important role in development of vertebrate embryos. We demonstrate impacts of exogenous RA on the formation of neural crest cells in zebrafish using specific neural crest markers sox9b and crestin. Treatment with all -trans RA at 10₋₇ mmol/L at 50% epiboly induces sox9b expression in the forebrain and crestin expression in the forebrain and midbrain, resulting in significant increase of pigment cells in the head derived from the cranial neural crest. In addition, RA treatment induces expression of sox9b and crestin in the caudal marginal cells of the neuroectoderm during early segmentation. Earlier commitment of these cells to the neural crest fate in the posterior margins leads to abnormal development of the posterior body, probably by preventing mingling of ventral derived and dorsal-derived cells during the formation of the tailbud.     

Structure and function of PBS model complex

A complex of phycobiliproteins, containing phycoerythrocyanin (PEC), C-phycovyanin (C-PC) and allo-phycocyanin (APC) as well as some linker polypeptides, was reconstructed. The absorption and fluoreacence spectra of the complex were compared with those of native phycobilisomes (PBS) and the phycobiliproteins. Based on the measured data, it can be concluded that the complex can be taken as a model of PBS and is an entirely functional group for excitation energy transfer step by step from peripheral PEC to APC. The single terminal emitter feature of the complex makes it favorable for clarifying energy transfer pathways and the kinetics in comparison with native PBS. Further research is carried on in the lab.     

基于RBF神经网络的复杂系统的建模与优化

针对BP神经网络算法在用于函数逼近时,存在着收敛速度慢、易陷入局部极小的不足,提出基于RBF(径向基函数,Radail Basis Function)神经网络的建模与优化方法,并以典型复杂系统联合制碱工业过程为例,利用神经网络算法的强大学习能力建立RBF神经网络模型,并进行优化研究。以联合制碱工业过程中的煅烧工段为例进行了仿真研究,仿真结果显示RBF神经网络的优越性,效果令人满意。     

基于FFT和神经网络的复模态参数识别

为了精确识别结构复模态参数,提出了一种基于快速傅里叶变换(FFT)和人工神经网络的模态识别方法.该方法首先对自由振动信号进行FFT预处理,得到粗略的各阶模态频率和相位.然后,根据模态的阶数设定神经元的个数,根据预处理后得到的频率和相位设定神经网络权值和基函数参数迭代的初始值.最后,通过对人工神经网络进行训练,达到利用自由振动信号进行时域模态识别的目的.仿真结果表明,该算法可消除频率法识别中因频谱泄露与噪声等产生的误差,提高模态识别的精度,因而是一种有效的时域识别方法.     

Syn-deformationP-T paths of Xiaoqinling metamorphic core complex

TheP-T paths of the Xiaoqinling metamorphic core complex (XMCC) have been investigated with the Gibbs method by researching the compositional changing of zoned epidotes which formed during syn-deformation metamorphism. TheseP-T paths indicate that the XMCC had experienced the following thermodynamics processes: firstly, near isobaric falling slightly in temperature in lower crust; secondly, fast decompression and rising in temperature during extensional uplifting to middle crust level; and finally, isobaric falling in temperature in middle crust. The upwelling and emplacement of the deep magma may be a major factor during the uplifting processes of the metamorphic core complex.     

Expression of neurofilament gene in spinal motoneurons during neural regeneration

The techniquein situ hybridization was used to measure the levels of light (NF-L), medium (NF-M) and heavy (NF-H) neurofilament protein subunits mRNA in L_4-6 spinal motoneurons in adult rat during regeneration following a unilateral crush of the sciatic nerve. It was found that the hybridization signals of each neurofilament subunit rnRNA were dramatically decreased in spinal motoneurons postaxotomy by light microscopy. The hybridization signals of NF-L and NF-M mRNA were located in cytoplasm of neurons, whereas NF-H mRNA was found in both nucleus and cytoplasm of neurons. Image analysis showed that the encoding levels of mRNA for each of neurofilament subunit mRNA reduced on the 3rd d and returned to control levels on the 28th d following the lesion. The relative levels of mRNA coding for each neurofilament subunit were significantly different. The lowest level of NF-L mRNA was observed at 5 d postaxotomy, and that of NF-M, NF-H mRNA on the 7th and 10th d after injury. Moreover, the levels of HF-M and NF-H mRNA were reduced much lower and lasted much longer than that of NF-L mRNA. The observations suggest that there were different mechanisms for the regulation of neurofilament subunit genes expression. The reduced neurofilament gene expression may be due to a response to axonal injury and advance the restructure of axonal architecture.     

基于双树复数小波四元数卷积网络的Copy-move盲取证算法

随着图像编辑软件的普及与完善,使得人们通过Copy-move操作便可伪造图像,而现有的Copy-move盲取证算法很难提取到彩色图像的一致性特征,且结果依赖于手动调节参数,难以定位到准确的篡改区域.为此,利用四元数卷积网络提取彩色图像空间一致性信息和双树复数小波提取图像局部信息的优势,提出了一种基于双树复数小波四元数卷...     

复杂背景下结合颜色和分形特征的多目标检测

针对复杂自然背景下的多目标检测,提出了结合颜色和分形特征的多目标检测算法.将RGB颜色空间转换到Lab颜色空间,采用改进K-means聚类算法,去除大片背景区域,计算区域分形维数和分形拟合误差.两种分形特征相结合能够准确排除小面积背景奇异区域的干扰,检测出待测图像中的多个目标.仿真结果表明:该算法能够正确检测出复杂自然背景下的多个目标,对彩色图像分割后的保留区域求分形特征,避免了搜索目标带来的计算量.相比于对全图提取分形特征的方法,本算法在时间上缩短约80%.     

基于复Morlet变换和改进AlexNet神经网络的柴油机气门间隙异常故障诊断方法

针对柴油机缸盖振动信号非线性、非平稳的特点,以及传统故障诊断方法需要先验知识且特征提取费时费力的缺点,提出了一种基于复Morlet变换和改进AlexNet神经网络的柴油机气门间隙异常故障诊断方法。首先通过复Morlet小波将柴油机缸盖振动信号转换为时频图,该变换包含了信号的时频域信息,比单一的时域或频域信号更适合分析柴油机缸盖振动这种非平稳信号;其次将时频图输入至AlexNet神经网络进行特征自动提取并建立故障诊断模型,解决了传统手工提取特征费时费力且需要专家经验的问题;然后通过Batch Normalization和Dropout技术改进网络结构,并优化神经网络超参数以提高模型的准确度和计算效率;最后将本文方法与传统的故障诊断方法应用于柴油机气门间隙异常故障诊断并进行对比,发现其诊断准确率最高,验证了所提方法的优越性。     

不同水分条件对3种蕨类植物配子体形态发育的影响

研究了不同水分条件对3种不同演化地位的蕨类植物(海金沙(Lygodium japonicum)、水蕨(Ceratopteris thalictroides)、狗脊(Woodwardia japonica))的配子体形态发育的影响.结果表明:1)随着水分的增加,海金沙、水蕨和狗脊孢子的萌发速率加快,丝状体的长度增长; 2)在缺水条件下,配子体易发育成雄性配子体,而在水分充足条件下,配子体易发育成两性配子体或雌性配子体; 3)次生假根的形态及毛状体的数量等受到水分条件的影响; 4)孢子萌发类型、配子体发育类型及成熟原叶体和毛状体的形态结构是相对稳定的性状,可作为蕨类植物分类鉴定的依据.     

激素和PEG对华重楼种子萌发的影响

为研究不同浓度的赤霉素(GA3)、2,4-二氯苯氧乙酸(2,4-D)和不同浓度的PEG4000、PEG6000、PEG8000和PEG10000对华重楼种子萌发的影响,采用培养皿培养法,用不同浓度的2种激素和4种PEG浸种处理,观察记录了种子萌发率和萌发开始时间(胚根突破种皮所需的天数)、第120 d时胚根长3 cm的种子所占的比例,筛选了促进华重楼种子萌发的最适浸种试剂和浓度.结果表明:500 mg/L GA3,0.5 mg/L 2,4-D,25%(质量比)PEG4000和25%PEG8000浸种能有效打破华重楼种子休眠并促进种子萌发,其中500 mg/L GA3效果最佳.这是国内首次关于华重楼种子萌发条件的全面研究报道.     

The development of embryo and endosperm inRanalisma rostratum

The embryo and endosperm development inRanalisma rostratum was studied in this paper. The zygote divides by a transverse wall to form basal and apical cells. The larger basal cell undergoes no further division and becomes the basal suspensor cell. The embryo development conforms to the Caryophyllad type. In the mature seed, the embryo is U-shaped and forms the embryonic shoot apex accompanied by one intraseminal leaf. The endosperm development belongs to the Hebolial type. The primary endosperm nucleus (invariably lying in the chalazal part of the embryo sac) divides and forms a large micropylar chamber and a small chalazal chamber. The chalazal endosperm chamber remains uninucleate, and free nuclear divisions occur in the micropylar chamber. Cellularization of the endosperm begins simultaneously from the micropylar and the chalazal ends. The cellularization is related to phragmoplast and cell plate, and does not arise from the free growth wall. Accompanying the embryo development, the endosperm degrades and does not exist in the mature seed. In the proembryo developing stages, the basal suspensor cell and chalazal endosperm cell grow to a considerable size. The nucleus and cytoplasm stain densely. Sometimes, binucleate or trinucleate was observed in both of the two cells. In the laterally concave embryo stage, they begin to degenerate which coincides with the cellularization of the endosperm. They are presumed to participate in the nourishment of the embryo proper in the early stages of embryo development. Supported by the National Natural Science Foundation of China Wang Jianbo: born in May. 1964, Ph. D.     

红树蚬与歪红树蚬形态差异分析

【目的】从形态上区分红树蚬(Polymesoda erosa)和歪红树蚬(Polymesoda expansa)。【方法】运用多变量形态度量学分析方法,采用14项生物学形态性状指标比较广西廉州湾草头村红树林的红树蚬与歪红树蚬的差异。【结果】方差分析结果显示,1个比例性状达到差异显著水平(P0.05),12个比例性状达到差异极显著水平(P0.01)。主成分分析构建了4个主成分,第1主成分贡献率为30.70%,第2主成分贡献率为17.76%,第3主成分贡献率为10.11%,第4主成分贡献率为7.80%,累计贡献率为66.37%。逐步判别分析建立判别函数,其判别准确率p1为81.7%~89.0%,p2为81.7%~87.5%,综合判别率为86.0%。【结论】壳高、壳宽、韧带长、韧带宽等性状决定红树蚬和歪红树蚬两个种间的形态差异,判别方程可以有效将上述种群区分开来。     

橙皮苷铜配合反应的工艺研究

以橙皮苷和铜金属盐为原料进行配合反应合成橙皮苷-铜配合物.以产率为指标.通过单因素和正交试验确定了合成橙皮苷-铜配合物的最佳条件为pH值8.5,反应温度45℃,反应时间45 min,所得配合物产率达72.08%.且配合物在水中溶解度为1800μg/mL,是原橙皮苷溶解度的100倍左右.     

矩阵的运算与K-复数的运算关系

矩阵与K-复数之间存在一定联系.在矩阵与K-复数理论基础上,讨论了矩阵转置、对称方阵、反对称矩阵、矩阵的分解、方阵的行列式及可逆矩阵与K-复数的关系.获得K-复数用矩阵表示后,它的运算可以转化为矩阵的运算.所得结果是复数中相应结果的应用.     

窃听双向协作网络机会中继选择方案的保密性能分析

对于多中继、窃听双向协作网络,研究使用机会中继选择和模拟网络编码技术来增强窃听系统物理层安全性.具体地,窃听双向协作网络的通信过程共3个时隙.第1和第2时隙,两信源分别发送消息到中继节点;第3时隙,根据最大最小中继选择方案,选取一个成功解码两信源消息的中继节点.使用模拟网络编码技术将两信源消息混合,把混合消息广播给两信...     

鱼用催产激素在倒刺鲃属(Spinibarbus)鱼类人工繁殖中的应用进展

倒刺鲃属(Spinibarbus)鱼类是我国地方特色鱼类,其人工繁殖必须借助催产激素。近年来,倒刺鲃属鱼类鱼苗需求激增,供不应求。本文从神经系统和内分泌激素对鱼类性腺发育成熟的调节、倒刺鲃属鱼类的繁殖习性、鱼用催产激素种类、倒刺鲃属鱼类催产激素和使用剂量、催产激素的使用方式和配制等方面综述了鱼用催产激素在倒刺鲃属鱼类人工繁殖中的应用进展,以期为倒刺鲃属鱼类规模化生产提供理论基础和实践方案。     

披针新月蕨的配子体发育及卵发生

采用显微观察技术对披针新月蕨（Pronephrium penangianum）的配子体发育和卵发生进行了研究.结果表明：披针新月蕨孢子褐色,二面型,接种后约12 d,孢子萌发,经丝状体和片状体阶段,发育成原叶体,孢子萌发是书带蕨型（Vittaria-type）,原叶体发育是槲蕨型（Drynaria-type）,成熟原叶体是心形,接种46 d后产生颈卵器和精子器.披针新月蕨的颈卵器起源于原叶体生长点下方的原始细胞.原始细胞经过2次不等分裂,形成3个细胞,分别是顶细胞、初生细胞和基细胞.其中,初生细胞不等分裂,形成中央细胞和颈沟细胞.随后,中央细胞不等分裂形成腹沟细胞和卵细胞.随着卵细胞成熟,颈沟细胞和腹沟细胞逐渐退化,但腹沟细胞通过孔区与卵细胞相连.卵细胞成熟后,细胞上表面具有明显的卵膜,中央具有受精孔.本研究对于完善蕨类植物生殖生物学和探讨蕨类植物的演化具有重要意义.     

苦楝树春季物候期特征及其与农业界限温度关系

根据广西 12个市县气象站对苦楝物候观测资料 (其中 2 4°N以北 4个点 ,2 4°N和以南 8个点 ,资料年限14年～ 19年 ) ,发现 ,各站苦楝春季平均芽开放期 (y芽开放 )、展叶平均始期 (y展叶始 )、开花平均始期 (y开花始 )与当地经度、纬度、冬季平均气温呈正相关关系 (r芽开放 =0 .85 * * ,r展叶始 =0 .86 * * ,r开花始 =0 .99* * * )。在广西 2 4°N及以南地区 ,用各站逐年 1月平均气温 (X1 1 )、≥ 10℃始日 (X1 2 )和头年≥ 15℃终日 (X1 3 )建立的苦楝物候期预测模型 :y1 芽开放 =87.995 - 0 .2 6 3X1 1 + 0 .16 3X1 2 - 0 .12 3X1 3(r =0 .5 5 * * * ,n =130 ) ;y1 展叶始 =74.95 3 - 0 .35 8X1 1+ 0 .2 15 X1 2 + 0 .111X1 3 (r =0 .70 * * *,n =130 ) ;y1 开花始 =46 .0 6 3 - 0 .2 6 7X1 1 + 0 .2 5 0 X1 2 + 0 .0 6 5 X1 3(r =0 .71* * * ,n =130 )。在广西 2 4°N以北地区 ,用各站逐年≥ 5℃初日 (X2 1 )、1月平均气温 (X2 2 )和头年≥ 10℃终日 (X2 3)建立的苦楝物候期预测模型 :y2芽开放 =87.2 0 4+ 0 .12 3X2 1 - 0 .2 16 X2 2 - 0 .0 5 X2 3(r=0 .5 0 * * ,n =6 9) ;y2展叶始 =80 .5 72 + 0 .12 6 X2 1 - 0 .334X2 2 - 0 .142 X2 3(r =0 .6 4* * *,n =6 9) ;y2开花始 =89.