基于快速神经网络架构搜索的鲁棒图像水印网络算法

为解决深度学习在图像水印算法中计算量大且模型冗余的问题，提高图像水印算法在抵抗噪声、旋转和剪裁等攻击时的鲁棒性，提出基于快速神经网络架构搜索（neural architecture search，NAS）的鲁棒图像水印网络算法。通过多项式分布学习快速神经网络架构搜索算法，在预设的搜索空间中搜索最优网络结构，进行图像水印的高效嵌入与鲁棒提取。首先，将子网络中线性连接的全卷积层设置为独立的神经单元结构，并参数化表示结构单元内节点的连接，预先设定结构单元内每个神经元操作的搜索空间；其次，在完成一个批次的数据集训练后，依据神经元操作中的被采样次数和平均损失函数值动态更新概率；最后，重新训练搜索完成的网络。水印网络模型的参数量较原始网络模型缩减了92%以上，大大缩短了模型训练时间。由于搜索得到的网络结构更为紧凑，本文算法具有较高的时间性能和较好的实验效果，在隐藏图像时，对空域信息的依赖比原始网络更少。对改进前后的2个网络进行了大量鲁棒性实验，对比发现，本文算法在CIFAR-10数据集上对抵抗椒盐噪声和旋转、移除像素行（列）等攻击优势显著；在ImageNet数据集上对抵抗椒盐高斯噪声、旋转、中值滤波、高斯滤波、JPEG压缩、裁剪等攻击优势显著，特别是对随机移除行（列）和椒盐噪声有较强的鲁棒性。     

Forthrightly monitoring ferroptosis induced by endoplasmic reticulum stresses through fluorescence lifetime imaging of microviscosity increases with a specific rotor

To test the hypothesis that the microviscosity changes of Endoplasmic Reticulum(ER) can be a useful indicator of ferroptosis promoted by ER Stresses(ERS), a new ER targeting viscosity rotor, L-Vis-1 was developed and applied in the quantitation of viscosity by FLIM imaging in live cells. The FLIM imaging exhibited an excellent resolution almost as good as the corresponding confocal imaging, more significantly, during ferroptosis processes promoted by different types of ERS, the viscosity increas...     

Numerical study of dynamic zigzag patterns in migrating epithelial tissue

Collective cell migration plays a crucial role in embryonic development, metastasis, and wound healing. Nevertheless, to the best of our knowledge, how the coordination between the cell motility and deformations affects the collective motion of epithelial cells is not fully understood. In this work, we propose a modified self-propelled Voronoi model for epithelial cell migration incorporating the coupling between the self-propulsion of cells and the polarization of the cell elongation. At a high coupling strength,we observe the emergence of backward traveling band structures formed by highly aligned cells, which can be regulated by cell elongations or shape anisotropy. Increasing the cell shape anisotropy, we find that large bands split into multiple small microbands. The latter essentially forms a dynamic zigzag pattern, in which the angle between the polarization direction of the bands and the migration direction switches alternatively between π/4 and-π/4 because the cells are forced to move preferentially in the anterior direction. We also analyzed the disclinations in the cell monolayer, force distribution near the domain boundaries and the shape alignment of the epithelial monolayer during the formation of this dynamic pattern. The present findings may further our understanding of stripe pattern formations in living systems and inspire potential designs for cell sorting.     

Cleavage/cross-coupling strategy for converting β-O-4 linkage lignin model compounds into high valued benzyl amines via dual C–O bond cleavage

Lignin is the most recalcitrant of the three components of lignocellulosic biomass. The strength and stability of the linkages have long been a great challenge for the degradation and valorization of lignin biomass to obtain bio-fuels and commercial chemicals. Up to now, the selective cleavage of C–O linkages of lignin to afford chemicals contains only C, H and O atoms. Our group has developed a cleavage/crosscoupling strategy for converting 4-O-5 linkage lignin model compounds into high value-a...     

基于MOF (Cu)@biochar气凝胶的制备及其在缓释氮肥上的应用

以羧甲基纤维素(CMC)、明胶和MOF(Cu)@biochar为原料,采用简单有效的冷冻干燥方法制备了(CMC/Gelatin/MOF(Cu)@biochar)杂化气凝胶,并用傅里叶变换红外光谱(FT-IR)、热重分析(TG)、X射线粉末衍射(XRD)、扫描电镜(SEM)和X射线光电子能谱(XPS)技术对其进行了表征;研究了MOF(Cu)@biochar含量、pH和不同的盐水溶液对杂化气凝胶溶胀行为的影响;以该气凝胶负载氯化铵,制备了一种新型缓释肥料(SRF),并研究了含2%(wt)SRF的沙性土壤的保水能力;SRF在土壤中第30天的累积释放率为79.4%;肥料在土壤中释放符合非Fickian扩散和阳离子交换的协同作用机理。     

化学类拔尖学生培养与高校青年教师成长协同发展实践*

在高等教育大众化的背景下,通过教学改革培养高素质的拔尖人才已显得非常重要。介绍一种以新入职的青年教师为实施主体的“理论课助教+实验课教学+科研训练”相结合的教学形式。通过连续、跟踪式的教学,不但可以选拔、培养出拔尖的化学类本科生,而且能够促进新入职的青年教师快速成长进步,起到“教学相长”的作用。     

突出矿井瓦斯抽放限制影响因素的AHP-熵权法评价模型

为提高煤与瓦斯突出矿井瓦斯抽放效果,建立了3个一级指标、14个二级指标的突出矿井瓦斯抽放限制影响因素评价指标体系,利用AHP和熵权法分别确定指标因子主、客观权重.通过实地调研分析和反馈验证了AHP-熵权法的可行性和正确性,利用加权平均法确定评价模型的综合权重.研究表明:封孔方式、钻孔半径、抽放时间、煤体裂隙发育程度和抽放负压是目前影响煤矿瓦斯抽放效果的主控因素.     

大尺寸电阻加热式碳化硅晶体生长热场设计与优化

大尺寸低缺陷碳化硅(SiC)单晶体是功率器件和射频(RF)器件的重要基础材料,物理气相传输(physical vapor transport, PVT)法是目前生长大尺寸SiC单晶体的主要方法。获得大尺寸高品质晶体的核心是通过调节组分、温度、压力实现气相组分在晶体生长界面均匀定向结晶,同时尽可能减小晶体的热应力。本文对电阻加热式8英寸(1英寸=2.54 cm)碳化硅大尺寸晶体生长系统展开热场设计研究。首先建立描述碳化硅原料受热分解热质输运及其多孔结构演变、系统热输运的物理和数学模型,进而使用数值模拟方法研究加热器位置、加热器功率和辐射孔径对温度分布的影响及其规律,并优化热场结构。数值模拟结果显示,通过优化散热孔形状、保温棉的结构等设计参数,电阻加热式大尺寸晶体生长系统在晶锭厚度变化、多孔介质原料消耗的情况下均能达到较低的晶体横向温度梯度和较高的纵向温度梯度。