中文分词与命名实体识别的联合学习

将卷积结构引入循环神经网络,从而构建卷积循环神经网络。以此为基础,研究构建了面向中文分词与实体识别联合学习的序列标注模型。该模型依托卷积循环神经网络构建特征编码层,实现中文字序列局部空间特征和长距离时序依赖特征的联合提取;依托改进的循环神经网络构建标签解码层,实现标签序列长距离时序依赖的有效建模;依托统一的分词与实体识别序列标注模式实现分词信息与实体信息的联合学习,避免传统流水线法的误差传播问题。在人民日报语料和微软标注语料上的实验结果显示,该框架较传统统计模型和神经网络模型有显著的性能提升,尤其是在识别字数较多的命名实体时,其效果明显优于其他方法。     

Adaptive subspace detection based on two-step dimension reduction in the underwater waveguide

In the underwater waveguide, the conventional adaptive subspace detector (ASD), derived by using the generalized likelihood ratio test (GLRT) theory, suffers from a significant degradation in detection per-formance when the samplings of training data are deficient. This paper proposes a dimension-reduced approach to alleviate this problem. The dimension reduction includes two steps: firstly, the full array is divided into several subarrays; secondly, the test data and the training data at each subarray are transformed into the modal domain from the hydrophone domain. Then the modal-domain test data and training data at each subarray are processed to formulate the subarray statistic by using the GLRT theory. The final test statistic of the dimension-reduced ASD (DR-ASD) is obtained by summing all the subarray statistics. After the dimension reduction, the unknown parameters can be estimated more accurately so the DR-ASD achieves a better detection performance than the ASD. In order to achieve the optimal detection performance, the processing gain of the DR-ASD is deduced to choose a proper number of subarrays. Simulation experiments verify the improved detection performance of the DR-ASD compared with the ASD.     

Simulation of the plasticizing behavior of composite modified double-base (CMDB) propellant in grooved calendar based on adaptive grid technology

The frequent occurrence of safety accidents during the calendering process is caused by the flammable and explosive properties of composite modified double-base (CMDB) propellant. Optimization of process parameters with the aid of fluid simulation technology could effectively ensure the safety of the calendering process. To improve the accuracy of the simulation results, material parameters and model structure were corrected based on actual conditions, and adaptive grid technology was applied in the local mesh refinement. In addition, the rheological behavior, motion trajectories and heat transfer mechanisms of CMDB propellant slurry were studied with different gaps, rotational rates and temperatures of two rollers. The results indicated that the refined mesh could significantly improve the contour clarity of boundaries and simulate the characteristics of CMDB propellant slurry reflux movement caused by the convergent flow near the outlet. Compared with the gap, the increased rotational rate of roller could promote the reflux movement and intensify the shear flow of slurry inside the flow region by viscous shear dragging. Meanwhile, under the synergistic effect of contact heat transfer as well as convective heat exchange, heat accumulated near the outlet and diffused along the reflux movement, which led to the countercurrent heat dissipation behavior of CMDB propellant slurry. The plasticizing mechanism of slurry and the safety of calendering under different conditions were explored, which provided theoretical guidance and reference data for the optimization of calendering process conditions. Based on the simulation results, the safety of the CMDB propellant calendering process could be significantly improved with a few tests conducted during a short research and development cycle.     

Finite-time tracking control and vibration suppression based on the concept of virtual control force for flexible two-link space robot

The dynamic modeling, finite-time trajectory tracking control and vibration suppression of a flexible two-link space robot are studied. Firstly, the dynamic model of the system is established by combining Lagrange method with assumed mode method. In order to ensure that the base attitude and the joints of space robot can reach the desired positions within a limited time, a non-singular fast terminal sliding mode (NFTSM) controller is designed, which realizes the finite-time convergence of the trajectory tracking errors. Subsequently, for the sake of suppressing the vibrations of flexible links, a hybrid tra-jectory based on the concept of the virtual control force is developed, which can reflect the flexible modes and the trajectory tracking errors simultaneously. By modifying the original control scheme, a NFTSM hybrid controller is proposed. The hybrid control scheme can not only realized attitude stabili-zation and trajectory tracking of joints in finite time, but also provide a new method of vibration sup-pression. The simulation results verify the effectiveness of the designed hybrid control strategy.     

Recursive weighted least squares estimation algorithm based on minimum model error principle

Kalman filter is commonly used in data filtering and parameters estimation of nonlinear system,such as projectile's trajectory estimation and control.While there is a drawback that the prior error covariance matrix and filter parameters are difficult to be determined,which may result in filtering divergence.As to the problem that the accuracy of state estimation for nonlinear ballistic model strongly depends on its mathematical model,we improve the weighted least squares method(WLSM)with minimum model error principle.Invariant embedding method is adopted to solve the cost function including the model error.With the knowledge of measurement data and measurement error covariance matrix,we use gradient descent algorithm to determine the weighting matrix of model error.The uncertainty and linearization error of model are recursively estimated by the proposed method,thus achieving an online filtering estimation of the observations.Simulation results indicate that the proposed recursive esti-mation algorithm is insensitive to initial conditions and of good robustness.     

支持原位计算的高效三角矩阵乘法向量化方法

向量化算法映射是向量处理器的难点问题。提出一种高效的支持原位计算的三角矩阵乘法向量化方法:将L1D配置为SRAM模式,用双缓冲的乒乓方式平滑多级存储结构的数据传输,使得内核的计算与DMA数据搬移完全重叠,让内核始终以峰值速度运行,从而取得最佳的计算效率;将不规则的三角矩阵乘法计算均衡分布到各个向量处理单元,充分开发向量处理器的多级并行性;将结果矩阵保存在乘数矩阵中,实现原位计算,节省了存储空间。实验结果表明,提出的向量化方法使三角矩阵乘法性能达到1053.7GFLOPS,效率为91.47%。     

机载叶簇穿透超宽带SAR系统关键指标设计

对影响叶簇穿透超宽带SAR系统性能的三个最主要的技术参量 (工作频率、极化形式和分辨率 )进行了深入系统的分析 ,其结论性意见对超宽带雷达系统顶层设计有实用参考价值     

决策支持系统可视化快速集成环境

决策支持系统的快速集成环境的研究一直是决策支持系统研究的核心内容。本文研究了决策支持系统可视化快速集成技术及其实现 ,并重点介绍了作者开发的决策支持系统可视化快速集成环境研究Vi sualDSSBuilder及其在开发过程中所解决的技术难题。     

基于图像信息的跟踪算法分析

图像跟踪算法研究是电视跟踪系统的关键技术之一.对波门跟踪算法、相关跟踪算法、多模跟踪算法、基于提边的差分跟踪算法及微分线性拟合外推跟踪算法进行了分析,并指出了电视跟踪系统跟踪算法的发展趋势及其在武器系统中的重要作用.     

武装直升机火箭武器对地攻击方式效能评估

根据某型武装直升机机载火箭武器的四种对地攻击方式,建立了该型直升机实战中将遭遇的作战条件的战场模型。采用层次分析法,建立了确定各作战条件模块加权系数的判断矩阵并计算了加权系数;结合各种对地攻击方式和该型直升机火箭武器系统的特点,确定了各攻击方式下各模块的效能系数。在加权系数和效能系数的基础上,计算了各作战条件下四种对地攻击方式各自的作战效能,得出一个当前作战条件下的最佳攻击方式提供给飞行员。为飞行员实施对地攻击提供了科学依据