基于多任务学习的多罪名案件信息联合抽取

面向法律文本的实体关系联合抽取技术对于案情关键信息的智能提取至关重要,是智慧司法领域应用中的重要环节。目前的联合抽取方法虽然已经在特定罪名案件的数据集上取得了较好的效果,但是由于模型在训练时只关注了特定罪名类型文本数据的特点,使得模型的泛化能力有限,在应用到多罪名案件的情况下常常使得模型的效果下降。因此引入多任务学习的方法对多罪名情形下的实体关系联合抽取进行了研究,以涉毒类案件和盗窃类案件两大类罪名的文书数据为基础,构建了一个罪名分类任务作为联合抽取的辅助任务,通过基于特征筛选的动态加权多任务模型同时对两个任务进行学习,在单任务模型的基础上整体F1值提升了2.4个百分点,在涉毒类案件和盗窃类案件上的F1值分别提升了1.6和3.2个百分点。     

A polarization modulated directional backlight autostereoscopic display

A polarization modulated directional backlight autostereoscopic display is proposed and demonstrated. The system consists of the orthogonally polarized backlight, the Fresnel lens array, a light shaping diffuser film, and a liquid crystal display (LCD) with a microphase retardation film. The autostereoscopic image pair carried by the directional light with different polarization directions is simultaneously projected to different spatial directions. The simulation and experimental results show that the directional projection of parallax images is realized for a high-quality autostereoscopic display with large viewing angle and continuous viewing volume, hence making it suitable for practical applications.     

航道艇无人智能测控系统技术研究

航道智能测量艇主要用于完成航道水下地形测量.无人智能测控系统由艇平台、感知、智能航控、岸基控制、通信等组成,可实现航路规划、自主循迹、远程操纵、数据传输、水下地形测绘、任务日志自动存档等功能.     

A Low-Complexity Precoding Scheme for Downlink Massive MU-MIMO Systems with Low-Resolution DACs

In this paper, we investigate the problem of downlink precoding for the narrowband massive multi-user multiple-input multiple-output (MU-MIMO) system with low-resolution digital-to-analog converters (DACs). We introduce a low-complexity precoding scheme based on the alternating direction method of multipliers (ADMM) framework in this work. An efficient gradient descent (GD) algorithm with adaptive step-size determination mechanism (ASGD) is proposed to alleviate the computational complexity bottleneck of the inherent matrix inversion. Numerical results demonstrate that the ASGD precoder achieves an attractive trade-off between the performance and computational complexity compared with other counterparts.

     

Focusing enhanced broadband metalens via height optimization

Metalenses are two-dimensional planar metamaterial lenses, which have the advantages of high efficiency and easy integration. Based on the method of spatial multiplexing, a metalens with a wide working waveband is designed by arranging TiO2 nanopillars under the resonance phase regulation. In addition, choosing an assistant metalens with optimized heights is effective to enhance metalens’s focusing, which is also illustrated in this paper. The metalens, designed with numerical aperture (NA) of 0.72 and center working wavelength of 600 nm, achieves the working waveband of 550—660 nm, the focus point’s size of below 420 nm, and the focusing efficiency of more than 30%.     

A survey on deploying mobile deep learning applications: A systemic and technical perspective

With the rapid development of mobile devices and deep learning, mobile smart applications using deep learning technology have sprung up. It satisfies multiple needs of users, network operators and service providers, and rapidly becomes a main research focus. In recent years, deep learning has achieved tremendous success in image processing, natural language processing, language analysis and other research fields. Despite the task performance has been greatly improved, the resources required to run these models have increased significantly. This poses a major challenge for deploying such applications on resource-restricted mobile devices. Mobile intelligence needs faster mobile processors, more storage space, smaller but more accurate models, and even the assistance of other network nodes. To help the readers establish a global concept of the entire research direction concisely, we classify the latest works in this field into two categories, which are local optimization on mobile devices and distributed optimization based on the computational position of machine learning tasks. We also list a few typical scenarios to make readers realize the importance and indispensability of mobile deep learning applications. Finally, we conjecture what the future may hold for deploying deep learning applications on mobile devices research, which may help to stimulate new ideas.     

一种高动态低信噪比环境下基于多样本点串行快速傅里叶变换的信号捕获方法

高超音速技术是未来空间飞行器的发展趋势,同时对通信平台在超高动态、低信噪比环境下的快速捕获能力也提出了新的挑战.针对经典捕获算法受频偏影响的局限性,该文提出一种基于信号多样本点串行快速傅里叶变换的信号捕获算法(MS-FFT),所提算法通过串行执行多个样本点的FFT,采用非相干合并后的峰值搜索得到捕获结果,在不增加复杂度的条件下,避免了频偏对捕获性能的影响.通过对峰值信噪比(PSNR)理论公式的推导,证明了MS-FFT的频偏适应范围取决于采样率,随着数模转换器件采样能力的不断提升,具有比经典算法更大的频偏适应范围.最后,通过仿真验证了上述理论推导的正确性,证明了所提算法更加适合超高动态环境的应用场景.     

Fast medical concept normalization for biomedical literature based on stack and index optimized self-attention

Neural Computing and Applications - Medical concept normalization aims to construct a semantic mapping between mentions and concepts and to uniformly represent mentions that belong to the same...     

Exploring Statistical Arbitrage Opportunities Using Machine Learning Strategy

Computational Economics - Arbitrage opportunity exploration is important to ensure the profitability of statistical arbitrage. Prior studies that concentrate on cointegration model and other...