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。     

Simultaneous determination of cinnamaldehyde and its metabolite in rat tissues by gas chromatography–mass spectrometry

Cinnamaldehyde (CA), an active ingredient isolated from the traditional Chinese medicine Cortex Cinnamomi, has a wide range of bioactivities. To clarify the distribution characteristics of CA, a selective and sensitive method utilizing gas chromatography–mass spetrometry was initially developed for simultaneously determining the concentration of CA and its metabolite cinnamyl alcohol in rat tissues. Selected ion masses of m/z 131, 105 and 92 were chosen, and separation of the analytes was performed on a DB‐5 ms (30 m × 0.25 mm, 0.25 µm, thickness) capillary column by gas chromatography–mass spectrometry. The calibration curves demonstrated good linearity and reproducibility over the range of 20–2000 and 20–4000 ng/mL for various tissue samples. Recoveries ranged from 86.8 to 107.5%, while intra‐ and interday relative standard deviations were all <11.3%. The analysis method was successfully applied in tissue distribution studies for CA and cinnamyl alcohol. As CA and cinnamyl alcohol may inter‐convert to one another, simultaneous determination of both analytes provides a comparative and accurate data for tissue study. The concentrations of CA and cinnamyl alcohol remaining in spleen were the highest among the main organs, including heart, liver, spleen, lung, kidney and brain. In addition, there was no long‐term accumulation of CA in rat tissues. Copyright © 2014 John Wiley & Sons, Ltd.     

Molecular design of ambipolar redox-active molecules II: closed-shell systems

This mini-review highlights key structural features that should be taken into account when creating ambipolar redox-active closed-shell metal-free molecules. This type of compound is strongly required for the fabrication of all-organic ‘poleless’ batteries and semiconductors. The suggested strategies aimed at stabilization of both oxidized (cationic) and reduced (anionic) redox-states are based on the comprehensive analysis of the most successful structures taken from the recent publications.     

A gold-based nanobeacon probe for fluorescence sensing of organophosphorus pesticides

A nanomaterials-based novel molecular beacon has attracted growing attentions in fluorescent assays as many nanomaterials possess excellent quenching efficiency. In this work, a gold-based nanobeacon probe was established to detect organophosphorus pesticides for the first time. The constructed gold-based nanobeacon acted as a signal indicator and could display the decreasing of the intensity in the presence of targets, which competitively bound to single strand DNA. To achieve a high sensitive probe, some parameters including solution pH, temperature and reaction time were investigated and optimized. The gold-based nanobeacon probe assay was proved to be rapid and sensitive to achieve a detection limit of 0.035 μM for isocarbophos, 0.134 μM for profenofos, 0.384 μM for phorate and 2.35 μM for omethoate, respectively. The prepared nanobeacon effectively reduced the background and improved the detection sensitivity and selectivity. The probe is stable, easy to operate and does not need sophisticated instruments. These features makes the probe feasible for screening trace organophosphorus pesticides in real samples.     

Metamorphic InAs(Sb)/InGaAs/InAlAs nanoheterostructures grown on GaAs for efficient mid-IR emitters

High-efficiency semiconductor lasers and light-emitting diodes operating in the 3–5?μm mid-infrared (mid-IR) spectral range are currently of great demand for a wide variety of applications, in particular, gas sensing, noninvasive medical tests, IR spectroscopy etc. III-V compounds with a lattice constant of about 6.1?Å are traditionally used for this spectral range. The attractive idea to fabricate such emitters on GaAs substrates by using In(Ga,Al)As compounds is restricted by either the minimum operating wavelength of ～8?μm in case of pseudomorphic AlGaAs-based quantum cascade lasers or requires utilization of thick metamorphic In_xAl_1-xAs buffer layers (MBLs) playing a key role in reducing the density of threading dislocations (TDs) in an active region, which otherwise result in a strong decay of the quantum efficiency of such mid-IR emitters. In this review we present the results of careful investigations of employing the convex-graded In_xAl_1-xAs MBLs for fabrication by molecular beam epitaxy on GaAs (001) substrates of In(Ga,Al)As heterostructures with a combined type-II/type-I InSb/InAs/InGaAs quantum well (QW) for efficient mid-IR emitters (3–3.6?μm). The issues of strain relaxation, elastic stress balance, efficiency of radiative and non-radiative recombination at T?=?10–300?K are discussed in relation to molecular beam epitaxy (MBE) growth conditions and designs of the structures. A wide complex of techniques including in-situ reflection high-energy electron diffraction, atomic force microscopy (AFM), scanning and transmission electron microscopies, X-ray diffractometry, reciprocal space mapping, selective area electron diffraction, as well as photoluminescence (PL) and Fourier-transformed infrared spectroscopy was used to study in detail structural and optical properties of the metamorphic QW structures. Optimization of the growth conditions (the substrate temperature, the As₄/III ratio) and elastic strain profiles governed by variation of an inverse step in the In content profile between the MBL and the InAlAs virtual substrate results in decrease in the TD density (down to 3?×?10⁷ cm^?2), increase of the thickness of the low-TD-density near-surface MBL region to 250–300?nm, the extremely low surface roughness with the RMS value of 1.6–2.4?nm, measured by AFM, as well as rather high 3.5?μm-PL intensity at temperatures up to 300?K in such structures. The obtained results indicate that the metamorphic InSb/In(Ga,Al)As QW heterostructures of proper design, grown under the optimum MBE conditions, are very promising for fabricating the efficient mid-IR emitters on a GaAs platform.     

First principles investigations of Fe,Co, Ni in model honeycomb carbon networks

The behaviors of ferromagnetic transition metals of the first period: Fe, Co and Ni are examined within density functional theory calculations in two dimensional carbon extended networks using model structure LiC₆. Around geometry optimized structures, the energy-volume equations of states considering non magnetic and spin polarized configurations established ferromagnetic ground states with magnetizations –reduced with respect to the metals’– of 2 μ_B for FeC₆ and 1 μ_B for CoC₆ while no magnetic solution could be identified for NiC₆. In the D_6h point group of the P6/mmm space group l_m decomposition of the d states results with increasing energy into doublet state E_1g with d(x²-y²) and d(xy); singlet state A_1g d(z²) and doublet state E_2g d(xz) and d(yz) lying on E_F and responsible of the onset of magnetic moments. This was mirrored via molecular orbital approach with a construct of Fe embedded between two extended carbon networks thus validating the model structure proposed for TC₆ compounds. The 100% polarization in one spin channel allows proposing potential uses in spintronics applications.     

Lithium Insertion Mechanism in Iron-Based Oxyfluorides with Anionic Vacancies Probed by PDF Analysis

The mechanism of lithium insertion that occurs in an iron oxyfluoride sample with a hexagonal–tungsten–bronze (HTB)-type structure was investigated by the pair distribution function. This study reveals that upon lithiation, the HTB framework collapses to yield disordered rutile and rock salt phases followed by a conversion reaction of the fluoride phase toward lithium fluoride and nanometer-sized metallic iron. The occurrence of anionic vacancies in the pristine framework was shown to strongly impact the electrochemical activity, that is, the reversible capacity scales with the content of anionic vacancies. Similar to FeOF-type electrodes, upon de-lithiation, a disordered rutile phase forms, showing that the anionic chemistry dictates the atomic arrangement of the re-oxidized phase. Finally, it was shown that the nanoscaling and structural rearrangement induced by the conversion reaction allow the in situ formation of new electrode materials with enhanced electrochemical properties.     

均匀磁场下碳离子笔形束的剂量变化分析及位置修正方法

分析外加均匀磁场对于碳离子笔形束剂量分布的影响,并考虑修正这种影响,为磁共振成像引导碳离子放射治疗的临床应用提供指导。本文利用蒙特卡罗方法模拟计算了不同能量碳离子笔形束在不同强度磁场下的剂量分布情况,发现垂直于碳离子束入射方向的均匀磁场对于碳离子笔形束射程缩短的影响很小,磁场对碳离子束的主要影响是引起束流横向偏转,特别是碳离子束布拉格峰位置的横向侧移。横向侧移程度与碳离子束的能量和磁场强度相关,根据模拟结果,得到了一个计算碳离子束布拉格峰在磁场中相对横向偏转的方程,并提出一种校正外加磁场引起的碳离子束布拉格峰横移的角度修正方法。这些结果可用于指导磁共振图像引导碳离子放射治疗计划系统的研发。     

An Improved Composite Multiscale Fuzzy Entropy for Feature Extraction of MI-EEG

Motor Imagery Electroencephalography (MI-EEG) has shown good prospects in neurorehabilitation, and the entropy-based nonlinear dynamic methods have been successfully applied to feature extraction of MI-EEG. Especially based on Multiscale Fuzzy Entropy (MFE), the fuzzy entropies of the τ coarse-grained sequences in τ scale are calculated and averaged to develop the Composite MFE (CMFE) with more feature information. However, the coarse-grained process fails to match the nonstationary characteristic of MI-EEG by a mean filtering algorithm. In this paper, CMFE is improved by assigning the different weight factors to the different sample points in the coarse-grained process, i.e., using the weighted mean filters instead of the original mean filters, which is conductive to signal filtering and feature extraction, and the resulting personalized Weighted CMFE (WCMFE) is more suitable to represent the nonstationary MI-EEG for different subjects. All the WCMFEs of multi-channel MI-EEG are fused in serial to construct the feature vector, which is evaluated by a back-propagation neural network. Based on a public dataset, extensive experiments are conducted, yielding a relatively higher classification accuracy by WCMFE, and the statistical significance is examined by two-sample t-test. The results suggest that WCMFE is superior to the other entropy-based and traditional feature extraction methods.     

o-Trityl phenoxy-imino vanadium (III) complexes: synthesis,characterization, and catalysis on ethylene (co)polymerization

Several phenoxy-imine ligands bearing o-trityl group in phenoxy moiety RN=CHArOH (Ar = C₆H₂(CPh₃)^tBu, R = 2,6-Me₂C₆H₃ ( L ₁ H ); 2,6-ⁱPr₂C₆H₃ ( L ₂ H ); 3,5-(CF₃)₂C₆H₃ ( L ₃ H ); 3,5-(OMe)₂C₆H₃ ( L ₄ H ); CHPh₂ ( L ₅ H ); CPh₃ ( L ₆ H )) were synthesized and characterized by¹H NMR and ¹³C NMR spectroscopy. The vanadium complexes based on these ligands LVCl₂(THF)₂ ( 1–6 ) were synthesized via conventional transmetalation reaction in moderate to high yields. Complexes 1–6 were fully characterized by FT-IR, elemental analyses and the molecular structures of 1 , 2 ·H₂O, (2 ·H₂O ) ₂ (μ-Cl) ₂ , 4 , and 5 were confirmed by X-ray crystallographic analysis in which the six-coordinated vanadium centers are in a typical octahedral geometry. Upon activation with Et₂AlCl in toluene, complexes 1–6 showed high activities in ethylene polymerization affording polymers with moderate molecular weight (5.9–11.8 × 10⁴ Da). Moreover, in hexane or CH₂Cl₂, 1–6 /Et₂AlCl exhibited enhanced activities. When activated with MAO or MMAO in toluene, these complexes showed relatively low activities but afforded polymers with ultra-high molecular weight (up to 3.30 × 10⁶ Da). 1–6 /Et₂AlCl also showed high activities in ethylene/1-hexene copolymerization at room temperature giving moderate molecular-weight polymers (6.5–11.4 × 10⁴ Da) with co-monomer incorporation being of 6.0 ~ 7.8%.