Structure learning on Bayesian networks by finding the optimal ordering with and without priors

Ordering based search methods have advantages over graph based search methods for structure learning of Bayesian networks in terms on the efficiency. With the aim of further increasing the accuracy of ordering based search methods, we first propose to increase the search space, which can facilitate escaping from the local optima. We present our search operators with majorizations, which are easy to implement. Experiments show that the proposed algorithm can obtain significantly more accurate results. With regard to the problem of the decrease on efficiency due to the increase of the search space, we then propose to add path priors as constraints into the swap process. We analyze the coefficient which may influence the performance of the proposed algorithm, the experiments show that the constraints can enhance the efficiency greatly, while has little effect on the accuracy. The final experiments show that, compared to other competitive methods, the proposed algorithm can find better solutions while holding high efficiency at the same time on both synthetic and real data sets.

Brain networks modeling for studying the mechanism underlying the development of Alzheimer’s disease

Alzheimer’s disease is a primary age-related neurodegenerative disorder that can result in impaired cognitive and memory functions.Although connections between changes in brain networks of Alzheimer’s disease patients have been established,the mechanisms that drive these alterations remain incompletely understood.This study,which was conducted in 2018 at Northeastern University in China,included data from 97 participants of the Alzheimer’s Disease Neuroimaging Initiative(ADNI)dataset covering genetics,imaging,and clinical data.All participants were divided into two groups:normal control(n=52;20 males and 32 females;mean age 73.90±4.72 years)and Alzheimer’s disease(n=45,23 males and 22 females;mean age 74.85±5.66).To uncover the wiring mechanisms that shaped changes in the topology of human brain networks of Alzheimer’s disease patients,we proposed a local naive Bayes brain network model based on graph theory.Our results showed that the proposed model provided an excellent fit to observe networks in all properties examined,including clustering coefficient,modularity,characteristic path length,network efficiency,betweenness,and degree distribution compared with empirical methods.This proposed model simulated the wiring changes in human brain networks between controls and Alzheimer’s disease patients.Our results demonstrate its utility in understanding relationships between brain tissue structure and cognitive or behavioral functions.The ADNI was performed in accordance with the Good Clinical Practice guidelines,US 21 CFR Part 50-Protection of Human Subjects,and Part 56-Institutional Review Boards(IRBs)/Research Good Clinical Practice guidelines Institutional Review Boards(IRBs)/Research Ethics Boards(REBs).

THE TOPOLOGICAL OPTIMIZATION FOR TRUSS STRUCTURES WITH STRESS CONSTRAINTS BASED ON THE EXIST-NULL COMBINED MODEL

A new exist-null combined model is proposed for the structural topology optimization. The model is applied to the topology optimization of the truss with stress constraints. Satisfactory computational result can be obtained with more rapid and more stable convergence as compared with the cross-sectional optimization. This work also shows that the presence of independent and continuous topological variable motivates the research of structural topology optimization.

工业过程关键指标预测的知识协同进化增强图卷积网络方法

在流程工业关键变量预测领域,已有研究致力于将过程知识与大数据相结合,以实现更高的准确性,降低过拟合风险和提高可解释性.然而,现有工作存在准确的先验知识构建成本高、无法从丰富的数据中挖掘知识等问题,限制了这些方法在实际工业过程中的广泛应用.为了解决这些挑战,本文提出了一种基于知识协同进化的增强图卷积网络方法.首先,利用易获取的过程流图构建低成本的粗粒度流程知识.然后,在图卷积神经网络模型训练中引入图探索,实现知识更新.最后,为了降低知识复杂度并保持一致性,设计了一种知识过滤机制.所提出的方法在基准的脱丁烷塔工艺过程上进行了验证.实验结果表明,该方法具有出色的预测准确性,并获得高质量的新知识.

基于课程学习权重集成的贝叶斯结构学习算法研究

从大量复杂的数据中学习贝叶斯网络(Bayesian network,BN)一直是一个难点问题,本文借鉴课程学习的思想,引入了一种适合于BN中节点之间互相影响程度的测量,然后划分课程阶段,分阶段构造无向图骨架,并利用优化函数对骨架进行优化;通过集成策略,将各个集成学习结果所得到的课程权重进行集合,并通过边过滤来减少错误边的出现;最后,通过爬山搜索构建BN结构。实验结果表明,在4个标准数据集上,本文所提方法具有较高的精确度和稳定性。与多种传统贝叶斯结构学习(Bayesian network structure learning,BNSL)方法相比,本文所提方法性能平均提高了37.18%。本文分析结果可为BNSL的增量学习过程进一步提供参考。

融合主题模型的图神经网络对话情感识别

对话情感识别(ERC)旨在预测对话中语句的情感类别。目前,基于图神经网络的ERC方法主要采用固定的超参数来确定图中边的连接,缺乏针对不同数据进行自适应构边的策略,且忽略了语句间的主题关系。此外,在图神经网络的训练过程中,这些方法通常采用求和叠加的方式来聚合节点信息,限制了模型的非线性能力。为此,本文将主题模型与图神经网络相融合,提出了一种新的构边方法。首先利用主题模型获取对话中语句的主题分布,然后将具有相同主题的语句相互连接。同时,引入了SwiGLU门控单元,用于调控图神经网络中层与层之间的信息流动。在边的类型方面,考虑了人物信息的差异,以更好地捕捉情感变化的内因和外因。通过在4个公开数据集(IEMOCAP、MELD、EmoryNLP、DailyDialogue)上进行的广泛实验,与当前先进的ERC方法相比,本文的方法在前3个数据集上的F1分数分别提升了1.69%,0.27%和0.38%。此外,本文的自适应方法在长对话上的效果提升了2.11%,优于短对话的0.8%,同时,通过引入SwiGLU有效减缓了图神经网络中的过度平滑现象。综合结果表明,本文提出的融合主题模型进行自适应构边以及引入SwiGLU门控单元的图神经网络方法,能够有效提高对话情感识别的效果,增强模型的鲁棒性。

增强分子拓扑信息的多任务图神经网络算法

以分子毒性为代表的分子属性预测在以药物设计为主的多个领域的发展中发挥着重要作用,但直接利用分子结构信息快速且准确地预测分子毒性一直是一个挑战。目前,卷积网络和图网络等深度学习方法的出现在这个问题的解决上得到了一定的进展。而以图网络为主的深度学习方法在分子毒性预测中存在两个关键问题,影响预测性能:第一,数据驱动使得模型在面对小批量数据时依然没有可靠的性能。第二,建模分子结构只考虑了天然共价键,只能提供粗粒度的信息。为解决上述问题,给出了一种对分子结构的新型建模方式MT-ToxGNN。该方法将多任务的思想融入图神经网络中,使得不同任务在训练时可以互相学习不同数据的可靠分布,从而避免在小批量数据上的过拟合问题。将分子编码成拓扑图结构时同时考虑分子内共价键以及非共价作用,就是在使用分子共价键构建传统图的边集之后,再使用非共价作用构建新型图的边集,从而弥补传统图网络对分子结构信息表示的不足。使用特别设计的图网络分别处理分子的共价与非共价信息,充分学习不同的分子结构。在与大量先进方法的性能比较中,MT-ToxGNN在多个分子毒性数据集上皮尔森系数指标达到了最佳。

基于图神经网络的固定骨架蛋白质设计方法研究

针对图神经网络(GNN)ProteinSolver结构特征约束不充分的问题,增加了骨架二面角、配对氨基酸的相对位置编码和相对方向等结构约束,提出了一种基于GNN的固定骨架蛋白质设计方法。实现了基于Transformer多头注意力机制的GNN架构,将物理坐标添加到消息传递和更新步骤中,提高了原子坐标的等变特性。在CATH数据集上的训练和测试结果显示:该文模型平均困惑度为8.12,比ProteinSolver的平均困惑度8.97降低了0.85;在掩盖率为50%时,ProteinSolver的恢复率为28.7%;然后,增加更多的结构约束,恢复率达到了30.3%;随后,将ProteinSolver的GNN替换成基于Transformer的GNN,恢复率达到了34.3%;最后,通过再引入等变特性,恢复率进一步提高到35.0%。

基于BERT-GAT-CorNet多标签中文短文本分类方法

多标签文本分类问题是多标签分类的一个重要内容,传统的多标签文本分类算法往往只关注文本本身的信息而无法理解深层语义信息,也未考虑标签之间的关系。为了解决这些问题,提出了融合BERT(Bidirectional Encoder Representation from Transformers)-GAT(Graph Attention neTwork)-CorNet(Correlation Network)的多标签文本分类模型。首先,通过预训练模型BERT表示文本的特征向量,并用生成的特征向量建立图结构数据;接着,用GAT来为不同节点分配不同的权重;最后,通过Softmax-CorNet学习标签相关性增强预测并分类。所提模型在今日头条子数据集(TNEWS)和KUAKE-QIC数据集上的准确率分别为93.3%和83.2%,通过对比实验表明,所提模型在多标签文本分类任务上性能得到了有效提升。

基于图卷积网络的柴油加氢生产软测量建模

精制柴油闪点和石脑油终馏点是柴油加氢生产最重要的质量指标,但目前缺乏测量这两种指标的在线分析仪表,软测量是可行的替代方案。为完成这两种指标的软测量建模,分析柴油加氢生产过程的工艺机理和过程数据特征,完成了辅助变量的重构和筛选,确定了两个软测量对象的辅助变量集,基于图卷积网络建立闪点和终馏点的软测量模型进行仿真研究。仿真研究结果表明:相比于目前应用广泛的软测量模型,所提出的软测量方法具有更高的精度,可以提升柴油加氢生产软测量建模的有效性和准确性。

Model-constrained and data-driven double-supervision acoustic impedance inversion

Seismic impedance inversion is an important technique for structure identification and reservoir prediction.Model-based and data-driven impedance inversion are the commonly used inversion methods.In practice,the geophysical inversion problem is essentially an ill-posedness problem,which means that there are many solutions corresponding to the same seismic data.Therefore,regularization schemes,which can provide stable and unique inversion results to some extent,have been introduced into the objective function as constrain terms.Among them,given a low-frequency initial impedance model is the most commonly used regularization method,which can provide a smooth and stable solution.However,this model-based inversion method relies heavily on the initial model and the inversion result is band limited to the effective frequency bandwidth of seismic data,which cannot effectively improve the seismic vertical resolution and is difficult to be applied to complex structural regions.Therefore,we propose a data-driven approach for high-resolution impedance inversion based on the bidirectional long short-term memory recurrent neural network,which regards seismic data as time-series rather than image-like patches.Compared with the model-based inversion method,the data-driven approach provides higher resolution inversion results,which demonstrates the effectiveness of the data-driven method for recovering the high-frequency components.However,judging from the inversion results for characterization the spatial distribution of thin-layer sands,the accuracy of high-frequency components is difficult to guarantee.Therefore,we add the model constraint to the objective function to overcome the shortages of relying only on the data-driven schemes.First,constructing the supervisor1 based on the bidirectional long short-term memory recurrent neural network,which provides the predicted impedance with higher resolution.Then,convolution constraint as supervisor2 is introduced into the objective function to guarantee the reliability and accuracy of the inversion results,which makes the synthetic seismic data obtained from the inversion result consistent with the input data.Finally,we test the proposed scheme based on the synthetic and field seismic data.Compared to model-based and purely data-driven impedance inversion methods,the proposed approach provides more accurate and reliable inversion results while with higher vertical resolution and better spatial continuity.The inversion results accurately characterize the spatial distribution relationship of thin sands.The model tests demonstrate that the model-constrained and data-driven impedance inversion scheme can effectively improve the thin-layer structure characterization based on the seismic data.Moreover,tests on the oil field data indicate the practicality and adaptability of the proposed method.

New approaches to quantify progressive damage and associated dynamic rock mass blockiness

In the past decade, numerical modelling has been increasingly used for simulating the mechanical behaviour of naturally fractured rock masses. In this paper, we introduce new algorithms for spatial and temporal analyses of newly generated fractures and blocks using an integrated discrete fracture network (DFN)-finite-discrete element method (FDEM) (DFN-FDEM) modelling approach. A fracture line calculator and analysis technique (i.e. discrete element method (DEM) fracture analysis, DEMFA) calculates the geometrical aspects of induced fractures using a dilation criterion. The resultant two-dimensional (2D) blocks are then identified and characterised using a graph structure. Block tracking trees allow track of newly generated blocks across timesteps and to analyse progressive breakage of these blocks into smaller blocks. Fracture statistics (number and total length of initial and induced fractures) are then related to the block forming processes to investigate damage evolution. The combination of various proposed methodologies together across various stages of modelling processes provides new insights to investigate the dependency of structure's resistance on the initial fracture configuration.

Feature Analysis and Modeling of the Network Community Structure

Community structure has an important influence on the structural and dynamic characteristics of the complex systems.So it has attracted a large number of researchers.However,due to its complexity,the mechanism of action of the community structure is still not clear to this day.In this paper,some features of the community structure have been discussed.And a constraint model of the community has been deduced.This model is effective to identify the communities.And especially,it is effective to identify the overlapping nodes between the communities.Then a community detection algorithm,which has linear time complexity,is proposed based on this constraint model,a proposed node similarity model and the Modularity Q.Through some experiments on a series of real-world and synthetic networks,the high performances of the algorithm and the constraint model have been illustrated.

图模数据解析下的配网单线图一键成图支持算法

为提升电力系统的自动配置能力,简化单线图操作过程,实现配网智能化调度,设计了图模数据解析下的配网单线图一键成图支持算法。以CIM(Common Information Model)模型为数据源,采用改进的肖维涅算法优化CIM数据质量,进行配网拓扑数据预处理,挖掘其所具有的关联性;通过转化拓扑关系形成特定树形,构建基于迭代二叉树3代算法的决策树机器学习模型,实现树节点坐标的自动分配;根据树边两端点坐标和该树边描述的设备种类,绘制矢量图形,简化单线图的操作过程。实验结果表明,该算法可有效提升配网图形的信息交互效果与电力系统自动配置能力,支持配网单线图一键成图功能。

周期运行图编制模型与算法研究

在周期运行的运输组织模式下,所有列车在车站到发都是周期循环发生的。将安排列车运行线的问题看作周期事件安排问题,并借助周期约束图及周期势差模型,可以建立周期运行图网络模型。模型充分考虑到列车不同情况下的停站时间、到发安全间隔等各项周期约束,并将列车的总停留时间最小作为目标函数。当约束图顶点和弧的数量众多时,模型的求解将比较困难。通过选择合适的约束图生成树,找到变量的合理取值范围,并对模型进行一些预先简化处理,可以降低模型的求解难度。最后求解一个区段不同列车开行方案的周期运行图,验证模型的可行性。

基于层次结构及模型驱动的配电网图形自动生成

随着配电网自动化及其相关系统建设水平的日益提高,系统之间的信息交换更加频繁。但鉴于当前不同系统对相同电力设备图元尚缺乏统一的定义标准,导致不同系统间图形缺乏交互能力,难以达到美观、一致性、实用化的需求。立足于配电网应用,通过拓扑模型抽取、对象识别、图元布局及智能布线等过程,实现了一种基于层次结构及模型驱动的配电网一次接线图的自动生成方法,为配电网自动化调度应用提供免维护的图形自动生成手段。通过工程现场的具体应用,验证了研究成果的有效性和实用性。

城市轨道交通网络周期运行图编制研究

将铺画城市轨道交通列车运行线的问题看作周期事件安排问题,分析周期约束特性、周期约束图等因素,考虑列车运行与停站时间、列车运行到发安全、列车折返和线间换乘衔接等周期约束,以减少平峰时段轨道交通路网乘客换乘的总等待时间为目标,建立城市轨道交通网络列车运行图编制模型;提出选择合适约束图生成树的方法,结合城市轨道交通运行特征和周期约束特性,对网络模型的基本圈约束进行细化分析,研究简化模型的方法;以北京城市轨道交通局部网络为实例,编制网络列车周期运行图。分析表明,模型及其优化算法可行,对城市轨道交通运行图编制工作有较强的实用性。

采用Stroke层次结构模型的道路网匹配方法

地理空间信息时刻处于变化之中,道路网变化尤其明显。如何确保道路数据的现势性成为重要研究课题。每条道路往往由多条弧段组成,目前的匹配算法大多对离散弧段单独匹配,而没有把每条道路作为一个整体进行匹配,导致即使是同一条道路,往往也存在局部匹配错误的情形。本文提出了基于Stroke层次结构模型的城市道路网匹配新方法。该方法采用Stroke技术把离散的道路弧段构建为完整的道路,亦即Stroke;进一步对Stroke进行分类分级,进而构建不同层次Stroke之间的关联匹配准则;最后采取由高等级Stroke到低等级Stroke逐级匹配、并以高等级Stroke约束较低等级Stroke匹配的方法与顺序,依次完成整个道路网之间的匹配过程。该算法的优势在于:把离散的道路弧段转化为整条Stroke进行整体匹配,避免了局部因素的影响;以高等级Stroke约束较低等级Stroke匹配,层层约束,算法运算效率与匹配正确率显著提高。

基于图论的直接连接热水供热系统热力工况计算模型

为解决集中供热管网热力失调问题和分析由于热网局部工况变化而引起的热力工况变化的规律和特点,基于图论方法,利用空间网络拓扑结构得到不同几何结构热网统一的热力工况计算模型。当热网工况发生变化时,运用该模型可对热网各热力参数进行计算和分析。

农民工的社会网络结构分析

采用2005年深圳外来农村流动人口的调查数据，利用指数随机图模型（p＊模型），分析了农民工的社会网络（包括社会支持网和社会讨论网）结构。分析结果显示，农民工社会网络关系稀疏，尤其是社会讨论网；无论在聚敛性还是扩张性方面，农民工社会网络的核心-边缘的局部结构均较明显，且有小团体现象产生；社会支持和社会讨论关系都更可能受到中间人的控制或约束。属性变量对社会支持网的影响较多，而对社会讨论网的影响较少。指数随机图模型为基于社会网络来认识农民工的社会化过程提供了新的方法。