Bulk Spin Torque-Driven Perpendicular Magnetization Switching in L10 FePt Single Layer

Due to its inherent superior perpendicular magnetocrystalline anisotropy, the FePt in L1₀ phase enables magnetic storage and memory devices with ultrahigh capacity. However, reversing the FePt magnetic state, and therefore encoding information, has proven to be extremely difficult. Here, it is demonstrated that an electric current can exert a large spin torque on an L1₀ FePt magnet, ultimately leading to reversible magnetization switching. The spin torque monotonically increases with increasing FePt thickness, exhibiting a bulk characteristic. Meanwhile, the spin torque effective fields and switching efficiency increase as the FePt approaches higher chemical ordering with stronger spin–orbit coupling. The symmetry breaking that generates spin torque within L1₀ FePt is shown to arise from an inherent structural gradient along the film normal direction. By artificially reversing the structural gradient, an opposite spin torque effect in L1₀ FePt is demonstrated. At last, the role of the disorder gradient in generating a substantial torque in a single ferromagnet is supported by theoretical calculations. These results will push forward the frontier of material systems for generating spin torques and will have a transformative impact on magnetic storage and spin memory devices with simple architecture, ultrahigh density, and readily application.     

Effect of Hf additions on phase transformation,microstructural stability,and hardness of nanocrystalline 304L stainless steels synthesized by mechanical alloying

304L stainless steels with Hf additions were nanostructured by mechanical alloying (MA) and annealed at temperatures up to 1100 °C. The results showed that face-centered cubic (fcc) phase in 304L transformed to body-centered cubic (bcc) phase during MA. The in-situ studies revealed that bcc-to-fcc phase transformation completed after 105 min annealing at 900 °C for 304L, whereas Hf addition increased the required time and temperature for the complete transformation. The grain size of 304L stainless steel was ~10 nm after MA and remained ~167 and ~293 nm after annealing at 900 and 1100 °C, respectively, with Hf addition in comparison to 960 nm average grain size of base 304L stainless steel after annealing at 900 °C. The hardness of 304L increased from ~200 HV to 408 HV after MA and remained 329 HV after annealing at 1100 °C with Hf addition as opposed to 195 HV hardness of 304L.     

Dual‐Locking Nanoparticles Disrupt the PD‐1/PD‐L1 Pathway for Efficient Cancer Immunotherapy

The clustered regularly interspaced short palindromic repeat (CRISPR)/CRISPR‐associated (Cas) enzyme, Cas13a, holds great promise in cancer treatment due to its potential for selective destruction of tumor cells via collateral effects after target recognition. However, these collateral effects do not specifically target tumor cells and may cause safety issues when administered systemically. Herein, a dual‐locking nanoparticle (DLNP) that can restrict CRISPR/Cas13a activation to tumor tissues is described. DLNP has a core–shell structure, in which the CRISPR/Cas13a system (plasmid DNA, pDNA) is encapsulated inside the core with a dual‐responsive polymer layer. This polymer layer endows the DLNP with enhanced stability during blood circulation or in normal tissues and facilitates cellular internalization of the CRISPR/Cas13a system and activation of gene editing upon entry into tumor tissue. After carefully screening and optimizing the CRISPR RNA (crRNA) sequence that targets programmed death‐ligand 1 (PD‐L1), DLNP demonstrates the effective activation of T‐cell‐mediated antitumor immunity and the reshaping of immunosuppressive tumor microenvironment (TME) in B16F10‐bearing mice, resulting in significantly enhanced antitumor effect and improved survival rate. Further development by replacing the specific crRNA of target genes can potentially make DLNP a universal platform for the rapid development of safe and efficient cancer immunotherapies.     

Investigation of assumptions and approximations in the virtual fields method for a viscoplastic material model

The Virtual Fields Method (VFM) is an inverse technique used for parameter estimation and calibration of constitutive models. Many assumptions and approximations—such as plane stress, incompressible plasticity, and spatial and temporal derivative calculations—are required to use VFM with full‐field deformation data, for example, from Digital Image Correlation (DIC). This work presents a comprehensive discussion of the effects of these assumptions and approximations on parameters identified by VFM for a viscoplastic material model for 304L stainless steel. We generated synthetic data from a Finite‐Element Analysis (FEA) in order to have a reference solution with a known material model and known model parameters, and we investigated four cases in which successively more assumptions and approximations were included in the data. We found that VFM is tolerant to small deviations from the plane stress condition in a small region of the sample, and that the incompressible plasticity assumption can be used to estimate thickness changes with little error. A local polynomial fit to the displacement data was successfully employed to compute the spatial displacement gradients. The choice of temporal derivative approximation (i.e., backwards difference versus central difference) was found to have a significant influence on the computed rate of deformation and on the VFM results for the rate‐dependent model used in this work. Finally, the noise introduced into the displacement data from a stereo‐DIC simulator was found to have negligible influence on the VFM results. Evaluating the effects of assumptions and approximations using synthetic data is a critical first step for verifying and validating VFM for specific applications. The results of this work provide the foundation for confidently using VFM for experimental data.     

编码曝光图像的L0正则化去模糊重建方法

目标的运动会导致其成像模糊。为了从模糊的图像中恢复清晰的目标图像,本文采用了编码曝光成像技术。与传统相机成像中快门一直处于开启的状态不同,编码曝光相机成像是在快门开启和闭合转换过程中成像。由于在时域快速转换的编码等效为频域较宽的滤波器,因此编码曝光成像有效地保留了目标的高频信息。为了从编码曝光图像中清晰地复原图像,本文设计了能保留图像高频细节的L ₀正则项约束的图像重建和模糊核估计方法。通过待重建图像与模糊核的交替迭代更新来完成图像重建。仿真合成图像和实际采集图像的实验表明,本方法对多种运动产生的模糊均有良好的图像复原效果。     

非贪婪的鲁棒性度量学习算法

目的 度量学习是机器学习与图像处理中依赖于任务的基础研究问题。由于实际应用背景复杂,在大量不可避免的噪声环境下,度量学习方法的性能受到一定影响。为了降低噪声影响,现有方法常用L1距离取代L2距离,这种方式可以同时减小相似样本和不相似样本的损失尺度,却忽略了噪声对类内和类间样本的不同影响。为此,本文提出了一种非贪婪的鲁棒性度量学习算法——基于L2/L1损失的边缘费歇尔分析（marginal Fisher analysis based on L2/L1 loss,MFA-L2/L1）,采用更具判别性的损失,可提升噪声环境下的识别性能。方法 在边缘费歇尔分析（marginal Fisher analysis,MFA）方法的基础上,所提模型采用L2距离刻画相似样本损失、L1距离刻画不相似样本损失,同时加大对两类样本的惩罚程度以提升方法的判别性。首先,针对模型非凸带来的求解困难,将目标函数转为迭代两个凸函数之差便于求解;然后,受DCA（difference of convex functions algorithm）思想启发,推导出非贪婪的迭代求解算法,求得最终度量矩阵;最后,算法的理论证明保证了迭代算法的收敛性。结果 在5个UCI（University of California Irrine）数据集和7个人脸数据集上进行对比实验：1）在不同程度噪声的5个UCI数据集上,MFA-L2/L1算法最优,且具有较好的抗噪性,尤其在30%噪声程度的Seeds和Wine数据集上,与次优方法LDA-NgL1（non-greedy L1-norm linear discriminant analysis））相比,MFA-L2/L1的准确率高出9%;2）在不同维度的AR和FEI人脸数据集上的实验,验证了模型采用L1损失、采用L2损失提升了模型的判别性;3）在Senthil、Yale、ORL、Caltech和UMIST人脸数据集的仿真实验中,MFA-L2/L1算法呈现出较强鲁棒性,性能排名第1。结论 本文提出了一种基于L2/L1损失的鲁棒性度量学习模型,并推导了一种便捷有效的非贪婪式求解算法,进行了算法收敛性的理论分析。在不同数据集的不同噪声情况下的实验结果表明,所提算法具有较好的识别率和鲁棒性。     

强干扰条件下机场道面细小裂缝自动识别算法

针对机场道面裂缝极其细小,而基于深度相机的裂缝检测技术面临道面表观结构复杂和平台剧烈震动的双重强干扰的难题,提出了结合L2正则化与动态阈值贪心策略的道面主轮廓建模算法,并基于此实现了机场道面毫米级细小裂缝的精确检测。首先,设计了基于L2正则化约束的道面主轮廓模型估计方法,解决了因表观结构复杂而导致的道面主轮廓过拟合问题;其次,提出基于动态阈值的改进贪心算法,通过迭代去除异常点的方式抑制检测平台震动带来的噪声干扰;最后,基于构建的道面主轮廓模型,提取并融合多方向的机场道面主轮廓,并利用裂缝的深度与形态信息实现裂缝提取。通过在真实机场道面数据集上的测试结果表明,该算法能够精确地完成道面主轮廓重建和细小裂缝识别,且识别性能优于多种现有经典的裂缝检测算法。     

Robust L1-norm two-dimensional collaborative representation-based projection for dimensionality reduction

Collaborative representation-based projection (CRP) is a well-known dimensionality reduction technique, which has been proved to have better performance than sparse representation-based projection (SRP) in the fields of recognition and computer vision. However, classical CRP is sensitive to noises and outliers since its objective function is based on L2-norm, and it will suffer from the curse of dimensionality as it is used for images processing. In this paper, a novel CRP model, named L1-norm two-dimensional collaborative representation-based projection (L1-2DCRP) and an efficient iterative algorithm to solve it are proposed. Different from conventional CRP, the optimal problem in our proposed model is a L1-norm-based maximization and the vector data is extended to matrix date. The proposed algorithm is theoretically proved to be monotonously convergent, and more robust to noises and outliers since L1-norm is used. Experimental results on CMU Multi-PIE, COIL20, FERET and ORL face databases validate the effectiveness of L1-2DCRP compared with several state-of-the-art approaches.     

An L0 regularized cartoon-texture decomposition model for restoring images corrupted by blur and impulse noise

In this paper, we address the image restoration case that includes both blurring and impulse noise. To recover an image with abundant features, we propose an L0 regularized cartoon-texture model for the simultaneous deblurring and impulse noise removal problem. We propose an L0 regularized framelet-based sparse representation and L0 regularized discrete cosine transform (DCT)-based sparse approximation to model the cartoon and texture of images, respectively. Unlike other cartoon-texture decomposition based-restoration approaches, our method does not depend on local features but globally controls the important non-zero components of the cartoon and texture in the framelet and DCT domain. Furthermore, we develop an alternating half-quadratic splitting method to solve the proposed L0 regularized cartoon-texture deblurring and impulse noise removal model (L0_RCTDINR) by introducing an alternating algorithm into the half-quadratic method. Experiments show the effectiveness of L0_RCTDINR on deblurring and impulse noise removal compared with existing state-of-the-art methods.     

认知无线多跳网络与以太网互联网关

为实现L2认知无线网络与以太网的互联,提出一种网络互联架构。为实现L2认知无线多跳网络中无线节点的快速寻址,设计一种编址方法,其利用IP地址和组网地址表征该网络中的无线节点;在网关中设计一种地址映射方案,通过设置地址映射表解决传统无线多跳网络中因广播造成的严重网络冗余和开销问题;为验证网关系统的通信能力、对传感数据的采集和处理能力,构建“智慧园区微环境数据采集”平台对该系统的有效性进行了验证。