基于GA_XGBoost模型的大学生科研能力预测问题研究

科学评价大学生科研创新能力对我国科研水平的提高具有重要意义.采用机器学习模型来预测大学生科研能力可以起到良好的效果,提出一种GAXGBoost模型来实现对大学生的科研能力预测.此模型是以Xgboost算法为基础,然后充分利用遗传算法的全局搜索能力自动搜索Xgboost最优超参数,避免了人为经验调参不准确的缺陷,最后采用精英选择策略以此确保每一轮都是最佳的进化结果.通过分析表明,所采用的GAXGBoost模型在大学生科研能力预测的结果中具有很高的精度,将此模型与Logistic Regression、Random Forest、SVM等模型进行对比,GAXGBoost模型的预测精度最高.     

Electrochemiluminescence Enhanced by the Synergetic Effect of Porphyrin and Multi-walled Carbon Nanotubes for Uric Acid Detection

The design and exploration of efficient, stable and environmentally compatible organic emitters for an electrochemiluminescence (ECL) sensor is a promising topic. Herein, a novel environmentally-friendly luminophore, ZnBCBTP@MWCNTs, were fabricated via self-assembly of porphyrin molecules (ZnBCBTP) onto multi-walled carbon nanotubes (MWCNTs). The resulting luminophore ZnBCBTP@MWCNTs displayed not only the highly ECL property and but also the good accelerated electron mobility. Then, a label-free ECL biosensor based ZnBCBTP@MWCNTs was constructed for the ultrasensitive detection of uric acid. Excitingly, this proposed ECL biosensor performed a good linear relationship in the range of 0–300 μM with a low detection limit of 1.4 μM, thus offering another reliable and feasible sensing platform for clinical bioanalysis with good selectivity, stability, and repeatability.     

The single-polarization filter composed of gold-coated photonic crystal fiber

A single-polarization filter comprising a gold-coated photonic crystal fiber based on surface plasmon resonance is designed and investigated. The pattern matching and coupled polarization characteristics analyzed by the full-vector finite element method (FEM) and losses at 1,540 nm are achieved to 1,016.01739 dB/cm (x-pol core mode) and 33.81917 dB/cm (y-pol core mode). The crosstalk (CT) value of the 1,540 nm band is ?853.12653 dB for fiber length $L=1,000\mathrm{\mu }\text{m}$ and the bandwidth is 850 nm. The working wavelength of the filter ranges from 1,280 nm to 1,540 nm by varying the diameter of outer air holes ($d_1$), the diameter of inner air holes ($d_4$), the metal film thickness (t), as well as the liquid refractive index (n).     

基于深度学习的ELM实时识别研究

基于深度学习的方法，在HL-2A装置上开发出了一套边缘局域模(ELM)实时识别算法。算法使用5200次放电数据(约24.19万数据切片)进行学习，得到一个深度为22层的卷积神经网络。为衡量算法的识别能力，识别了HL-2A装置自2009年实现稳定ELMy H模放电以来所有历史数据(约26000次放电数据)，共识别出1665次H模放电，其中误识别35次，误报率为2.10%。在实际的1634次H模放电中，漏识别4次，漏识别率为0.24%。该误报率和漏报率可以满足ELM实时识别的精度要求。识别算法在实时控制环境下，对单个时间点的平均计算时间为0.46ms，可以满足实时控制的计算速度要求。     

基于时变Copula模型的 基金收益率相依关系的应用研究

基于时变Copula模型,获得预测方差,确定单个基金收益率序列的边缘分布.利用常见的静态Copula和时变Copula模型对基金收益率序列间两两相依关系进行建模并进行对比分析.应用研究表明,基于MCMC方法的时变Copula模型能更有效地度量基金收益率序列的风险.     

激光能量对反钙钛矿GaCMn_3薄膜结构与物性的影响

本文采用脉冲激光沉积方法在LaAlO3(001)单晶衬底上制备了反钙钛矿GaCMn3薄膜,通过控制制备过程中脉冲激光的能量,研究了不同激光能量条件对GaCMn3薄膜结构与物理性能的影响.分别利用X射线衍射仪、原子力显微镜、超导量子干涉仪和物理性能测试系统,对所制备的薄膜的晶体结构、表面形貌和磁性、电输运性质进行了研究.结果表明,制备的样品均为具有多个晶面取向的反钙钛矿薄膜,且薄膜结构和物性明显随制备激光能量的变化而变化.当激光能量为450mJ时,制备的薄膜多晶面取向性最弱,结晶性和表面形貌最优良.实验所得的薄膜均表现出顺磁-铁磁-反铁磁相转变,然而转变过程比块材较平缓,同时薄膜的电阻率并未表现出块材中的突变特征,我们推测该现象很可能是由衬底的应力及衬底的晶格膨胀对薄膜反常晶格变化的抑制作用造成的.     

QEPAS系统中石英音叉模态仿真计算与研究

对石英音叉增强型光声光谱(QEPAS)系统中常用的石英音叉进行了有限元模态计算，获得石英音叉前6阶振型与模态频率，认知了第4阶对称摆动振型为有效振动，利用单因素法分析了石英音叉的音臂长度l₁、音臂宽度w₁、音臂厚度t、音臂切角θ、音臂圆孔直径d及音臂圆孔高度h对低阶有效共振频率(Fre)的影响，敏感度依次为： l₁> w₁>d>θ>t>h，考虑实际设计情形，筛选出了l₁，w₁，d与h四个石英音叉设计变量，采用Box-Behnken实验设计方案与RSM(response surface methodology)方法，以Fre为函数目标，建立l₁，w₁，d与h的二次回归响应面模型，得到了参数之间的交互作用，利用Design-Expert软件对响应面模型进行设计参数反求，结果表明，在15 000 Hz≤Fre≤25 000 Hz计算区域内误差较小，基本满足QEPAS系统的计算需求，所提出的研究与设计方法具有一定通用性，可为QEPAS系统中石英音叉结构参数设计提供参考。     

Studies on metabolism of total glucosides of paeony from Paeoniae Radix Alba in rats by UPLC‐Q‐TOF‐MS/MS

Total glucosides of paeony are the active constituents of Paeoniae Radix Alba. In this study, a novel strategy was proposed to find more metabolites and the differences between paeoniflorin, albiflorin and total glucosides of paeony (TGP). This strategy was characterized as follows: firstly, the animals were divided into three groups (paeoniflorin, albiflorin and TGP) to identify the source of TGP metabolites from paeoniflorin or albiflorin; secondly, a generic information‐dependent acquisition scan for the low‐level metabolites was triggered by the multiple mass defect filter and dynamic background subtraction; thirdly, the metabolites were identified with a combination of data‐processing methods including mass defect filtering, neutral loss filtering and product ion filtering; finally, a comparative study was used in the metabolism of paeoniflorin, albiflorin and TGP. Based on the strategy, 18 metabolites of TGP, 10 metabolites of paeoniflorin and 13 metabolites of albiflorin were identified respectively. The results indicated that the hydrolysis, conjugation reaction and oxidization were the major metabolic pathways, and the metabolic sites were the glycosidic linkage, the ester bond and the benzene ring. This study is first to explore the metabolism of TGP, and these findings enhance our understanding of the metabolism and the interactions of paeoniflrin and albiflorin in TGP. Copyright © 2015 John Wiley & Sons, Ltd.     

Numerical Evaluation of a Two-Loop Diagram in the Cutoff Regularization

The sunset diagram of λφ4 theory is evaluated numerically in cutoff scheme and a nonzero finite term(in accordance with dimensional regularization (DR) result) is found in contrast to published calculations. This findingdramatically reduces the critical couplings for symmetry breaking in the two-loop effective potential discussed in ourprevious work.