基于预训练语言模型的关键词感知问题生成

问题生成任务是指根据给定的文本段落和答案来自动生成对应的问题。针对现有问题生成方法存在的误差累积现象以及问题生成任务固有的“一对多”情况，提出一种带有关键词感知功能的问题生成方法。在预训练语言模型的基础上，实现关键词分类模型与问题生成模型的网络结构设计。输入文本段落中蕴含关键词，为使所生成的问题中包含同样的关键词以保证问题与段落的语义一致性，利用关键词分类模型提取出文本段落中的关键词，将关键词与非关键词的区分特征融入问题生成模型的输入中，该特征作为问题生成过程的全局信息，用以消除问题生成模型仅依赖局部最优解的弊端，减少误差累积与“一对多”情况的发生。在SQuAD数据集上的实验结果表明，该方法能够提升问题生成的质量，其BLEU-4指标值可达24，优于带有复制机制、带有语义监督的问题生成模型，目前已经借助百度百科数据平台实现了大规模工业应用。     

DEM: Deep Entity Matching Across Heterogeneous Information Networks

Heterogeneous information networks, which consist of multi-typed vertices representing objects and multi-typed edges representing relations between objects, are ubiquitous in the real world. In this paper, we study the problem of entity matching for heterogeneous information networks based on distributed network embedding and multi-layer perceptron with a highway network, and we propose a new method named DEM short for Deep Entity Matching. In contrast to the traditional entity matching methods, DEM utilizes the multi-layer perceptron with a highway network to explore the hidden relations to improve the performance of matching. Importantly, we incorporate DEM with the network embedding methodology, enabling highly efficient computing in a vectorized manner. DEM’s generic modeling of both the network structure and the entity attributes enables it to model various heterogeneous information networks flexibly. To illustrate its functionality, we apply the DEM algorithm to two real-world entity matching applications: user linkage under the social network analysis scenario that predicts the same or matched users in different social platforms and record linkage that predicts the same or matched records in different citation networks. Extensive experiments on real-world datasets demonstrate DEM’s effectiveness and rationality.

     

改进的多层感知机在客户流失预测中的应用

针对在传统的客户流失预测数据预处理中，使用one-hot编码处理离散属性导致数据维度增加及数据过于稀疏的问题，提出了两种基于多层感知机的改进后的客户流失预测模型。其主要思想是分别使用堆叠自编码器和实体嵌入两种方法对多层感知机进行改进，通过将离散属性的高维编码数据向低维空间映射，有效地减少了one-hot编码产生的稀疏数据，增加了离散属性值之间的关联度。在对两份公开的数据集进行交叉验证后的实验结果表明，改进后的模型既有效地提高了预测的准确度，又维持了传统多层感知机模型在并行化计算方面的优势。     

Embedding new observations via sparse-coding for non-linear manifold learning

Non-linear dimensionality reduction techniques are affected by two critical aspects: (i) the design of the adjacency graphs, and (ii) the embedding of new test data—the out-of-sample problem. For the first aspect, the proposed solutions, in general, were heuristically driven. For the second aspect, the difficulty resides in finding an accurate mapping that transfers unseen data samples into an existing manifold. Past works addressing these two aspects were heavily parametric in the sense that the optimal performance is only achieved for a suitable parameter choice that should be known in advance.     

Thumb push forces exertable by free-standing subjects

This study assessed the maximum pushing force that could be exerted by freestanding subjects, using thumbs only, on a small surface located in one of three horizontal positions at nine heights ranging from 13·5 to 169·9 cm above the floor. Each of 102 subjects made nine force exertion trials. Data are given on the cumulative percentages of subjects who could exert various forces at each of the 27 locations. An analysis of variance showed a highly significant effect due to location of the surface and highly significant interactions of location of the surface and trial presentation direction (top-down or bottom-up); of location of the surface and trial presentation pattern; and of location of the surface, trial presentation direction and trial presentation pattern. Maximum forces are exerted at about waist height and are increasingly reduced with increasing distances above and below the optimum location. Right-sided access to pushing surfaces is favoured over left-sided access.     

An Information Hiding Algorithm Based on Bitmap Resource of Portable Executable File

An information hiding algorithm is proposed, which hides information by embedding secret data into the palette of bitmap resources of portable executable （PE） files. This algorithm has higher security than some traditional ones because of integrating secret data and bitmap resources together. Through analyzing the principle of bitmap resources parsing in an operating system and the layer of resource data in PE files, a safe and useful solution is presented to solve two problems that bitmap resources are incorrectly analyzed and other resources data are confused in the process of data embedding. The feasibility and effectiveness of the proposed algorithm are confirmed through computer experiments.     

反应温度及时间对奥氏体不锈钢渗铬层组织结构的影响

利用固体粉末包埋技术对316H奥氏体不锈钢进行了1090℃保温0.5～20 h和750～1150℃保温10 h的化学热处理,研究了不同工艺参数对渗铬层组织结构和耐磨性的影响。通过光学金相显微镜(OM),扫描电镜(SEM),能谱仪(EDS)以及X射线衍射仪(XRD)等分析手段,研究了保温时间和反应温度对渗铬层的微观组织结构的影响规律,使用摩擦磨损试验机研究了不同工艺参数对渗铬层耐磨性的影响。结果表明:反应温度对渗铬层的厚度影响显著,在同一温度下进行渗铬,渗铬层的厚度与渗铬时间呈抛物线关系;渗铬层主要由Cr_(23)C_6、Cr_2C和α-Fe-Cr固溶体组成,在渗铬过程中,首先在试样表面形成碳铬化物层,随着反应温度提高或保温时间延长,在碳铬化合物层下会生成α-Fe-Cr固溶体层,进一步提高反应温度和保温时间,碳铬化合物层逐渐消失。渗铬后试样的耐磨性明显提高,且在相同温度下渗铬,保温时间短的耐磨性好;相同保温时间,反应温度低的耐磨性好。     

基于链路优先的快速协同虚拟网络映射算法

在网络虚拟化环境中,为了减少链路映射成本和提高收益,提出了一种链路优先的协同映射算法。该算法交替映射链路和节点,并使用贪心思想优先将带宽资源需求较大的虚拟链路映射到跳数较少的物理路径上,最终达到降低虚拟网络平均消耗和提高虚拟网络接受率的目的。仿真结果表明,与已有算法相比,所提算法节约了链路映射的成本,提高了虚拟网络接受率和物理网络收益,算法运行时间也有效缩短。     

一种提高嵌入率的语音融合隐藏改进算法

语音信号在较高的采样率下,采样数据具有冗余度。对于秘密语音信号,冗余度越大,算法的嵌入率越低。本文结合信号抽取插值方法,对已有的语音融合隐藏算法提出进一步的改进,以减小秘密语音的冗余度,提高算法的嵌入率。利用高质量的计算机网络这一信息传输途径,进行实验仿真,并从主客观两方面对算法的性能进行评估。实验结果表明,利用本文改进算法,秘密语音的嵌入率得到提高,且恢复的秘密语音信号质量较好,PESQ打分在4.0左右,可懂度高。算法具有较高的可行性。