Knowledge Graph Representation Learning Based on Automatic Network Search for Link Prediction

Link prediction,also known as Knowledge Graph Completion(KGC),is the common task in Knowledge Graphs(KGs)to predict missing connections between entities.Most existing methods focus on designing shallow,scalable models,which have less expressive than deep,multi-layer models.Furthermore,most operations like addition,matrix multiplications or factorization are handcrafted based on a few known relation patterns in several wellknown datasets,such as FB15k,WN18,etc.However,due to the diversity and complex nature of real-world data distribution,it is inherently difficult to preset all latent patterns.To address this issue,we proposeKGE-ANS,a novel knowledge graph embedding framework for general link prediction tasks using automatic network search.KGEANS can learn a deep,multi-layer effective architecture to adapt to different datasets through neural architecture search.In addition,the general search spacewe designed is tailored forKGtasks.We performextensive experiments on benchmark datasets and the dataset constructed in this paper.The results show that our KGE-ANS outperforms several state-of-the-art methods,especially on these datasets with complex relation patterns.

Improving the Syllable-Synchronous Network SearchAlgorithm for Word Decoding in ContinuousChinese Speech Recognition

The previously proposed syllable-synchronous network search (SSNS) algorithm plays a very important role in the word decoding of the continuous Chinese speech recognition and achieves satisfying performance. Several related key factors that may affect the overall word decoding effect are carefully studied in this paper, including the perfecting of the vocabulary, the big-discount Turing re-estimating of the N-Gram probabilities, and the managing of the searching path buffers. Based on these discussions, corresponding approaches to improving the SSNS algorithm are proposed. Compared with the previous version of SSNS algorithm, the new version decreases the Chinese character error rate (CCER) in the word decoding by 42.1% across a database consisting of a large number of testing sentences (syllable strings).

Enhancing the synchronizability of networks by rewiring based on tabu search and a local greedy algorithm

By considering the eigenratio of the Laplacian matrix as the synchronizability measure, this paper presents an efficient method to enhance the synchronizability of undirected and unweighted networks via rewiring. The rewiring method combines the use of tabu search and a local greedy algorithm so that an effective search of solutions can be achieved. As demonstrated in the simulation results, the performance of the proposed approach outperforms the existing methods for a large variety of initial networks, both in terms of speed and quality of solutions.

Structural reliability analysis using enhanced cuckoo search algorithm and artificial neural network

The present study proposed an enhanced cuckoo search(ECS) algorithm combined with artificial neural network(ANN) as the surrogate model to solve structural reliability problems. In order to enhance the accuracy and convergence rate of the original cuckoo search(CS) algorithm, the main parameters namely, abandon probability of worst nests paand search step sizeα0 are dynamically adjusted via nonlinear control equations. In addition, a global-best guided equation incorporating the information of global best nest is introduced to the ECS to enhance its exploitation. Then, the proposed ECS is linked to the well-trained ANN model for structural reliability analysis. The computational capability of the proposed algorithm is validated using five typical structural reliability problems and an engineering application. The comparison results show the efficiency and accuracy of the proposed algorithm.

Using Genetic Algorithms to Improve the Search of the Weight Space in Cascade－Correlation Neural Network

In this paper, we use the global search characteristics of genetic algorithms to help search the weight space of the neurons in the cascade-correlation architecture. The cascade-correlation learning architecture is a technique of training and building neural networks that starts with a simple network of neurons and adds additional neurons as they are needed to suit a particular problem. In our approach, instead ofmodifying the genetic algorithm to account for convergence problems, we search the weight-space using the genetic algorithm and then apply the gradient technique of Quickprop to optimize the weights. This hybrid algorithm which is a combination of genetic algorithms and cascade-correlation is applied to the two spirals problem. We also use our algorithm in the prediction of the cyclic oxidation resistance of Ni- and Co-base superalloys.

A Personalized Search Model Using Online Social Network Data Based on a Holonic Multiagent System

Personalized search utilizes user preferences to optimize search results,and most existing studies obtain user preferences by analyzing user behaviors in search engines that provide click-through data.However,the behavioral data are noisy because users often clicked some irrelevant documents to find their required information,and the new user cold start issue represents a serious problem,greatly reducing the performance of personalized search.This paper attempts to utilize online social network data to obtain user preferences that can be used to personalize search results,mine the knowledge of user interests,user influence and user relationships from online social networks,and use this knowledge to optimize the results returned by search engines.The proposed model is based on a holonic multiagent system that improves the adaptability and scalability of the model.The experimental results show that utilizing online social network data to implement personalized search is feasible and that online social network data are significant for personalized search.

Enhanced Energy Efficient Multipath Routing Protocol for Wireless Sensor Communication Networks Using Cuckoo Search Algorithm

Energy efficient routing is one of the major thrust areas in Wireless Sensor Communication Networks (WSCNs) and it attracts most of the researchers by its valuable applications and various challenges. Wireless sensor networks contain several nodes in its terrain region. Reducing the energy consumption over the WSCN has its significance since the nodes are battery powered. Various research methodologies were proposed by researchers in this area. One of the bio-inspired computing paradigms named Cuckoo search algorithm is used in this research work for finding the energy efficient path and routing is performed. Several performance metrics are taken into account for determining the performance of the proposed routing protocol such as throughput, packet delivery ratio, energy consumption and delay. Simulation is performed using NS2 and the results shows that the proposed routing protocol is better in terms of average throughput, and average energy consumption.

Cluster based hierarchical resource searching model in P2P network

For the problem of large network load generated by the Gnutella resource-searching model in Peer to Peer （P2P） network, a improved model to decrease the network expense is proposed, which establishes a duster in P2P network, auto-organizes logical layers, and applies a hybrid mechanism of directional searching and flooding. The performance analysis and simulation results show that the proposed hierarchical searching model has availably reduced the generated message load and that its searching-response time performance is as fairly good as that of the Gnutella model.

Hybrid ants-like search algorithms for P2P media streaming distribution in ad hoc networks

Media streaming delivery in wireless ad hoc networks is challenging due to the stringent resource restrictions,po-tential high loss rate and the decentralized architecture. To support long and high-quality streams,one viable approach is that a media stream is partitioned into segments,and then the segments are replicated in a network and served in a peer-to-peer(P2P) fashion. However,the searching strategy for segments is one key problem with the approach. This paper proposes a hybrid ants-like search algorithm(HASA) for P2P media streaming distribution in ad hoc networks. It takes the advantages of random walks and ants-like algorithms for searching in unstructured P2P networks,such as low transmitting latency,less jitter times,and low unnecessary traffic. We quantify the performance of our scheme in terms of response time,jitter times,and network messages for media streaming distribution. Simulation results showed that it can effectively improve the search efficiency for P2P media streaming distribution in ad hoc networks.

Greedysearch based service location in P2P networks

A model is built to analyze the performance of service location based on greedy search in P2P networks. Hops and relative QoS index of the node found in a service location process are used to evaluate the performance as well as the probability of locating the top 5% nodes with highest QoS level. Both model and simulation results show that, the performance of greedy search based service location improves significantly with the increase of the average degree of the network. It is found that, if changes of both overlay topology and QoS level of nodes can be ignored during a location process, greedy-search based service location has high probability of finding the nodes with relatively high QoS in small number of hops in a big overlay network. Model extension under arbitrary network degree distribution is also studied.

网络搜索数据与我国GDP的关联机理分析

网络搜索数据是研究我国宏观经济现象的重要微观信息依据。本文从需求、供给与政策三方面选取和筛选关键词合成网络搜索指数,并与我国GDP进行相关性研究。结果表明:网络搜索指数与GDP的相关性较高,且两者存在长期均衡关系与短期误差修正机制,当GDP逐渐偏离均衡,将会以1~2个月的调整速度从非均衡态过渡到均衡态;网络搜索指数的增长对我国GDP有促进作用。

SSA-MLP模型在岩质边坡稳定性预测中的应用

岩质边坡的力学参数量化及稳定性分析对岩质边坡灾害的防治具有重要意义。Hoek-Brown(H B)准则是一种用于确定岩体力学参数的经典方法,能反映出边坡岩体变形和位移的非线性破坏特征。在此基础上,首先,提出一种麻雀搜索算法(Sparrow Search Algorithm,SSA)改进多层感知器(Multi-Layer Perceptron,MLP)的神经网络模型,并用于边坡稳定性预测、指标敏感性分析及参数反演。其次,将收集的1085组岩质边坡的几何参数和H B准则参数等作为输入变量,极限平衡理论Bishop法求解的安全系数作为输出变量,对SSA MLP模型进行训练学习和性能评估。最后,将该模型运用于25个边坡实例,验证模型的有效性。结果显示,该模型收敛速度快、精度高,为边坡稳定性分析和参数量化提供了一种新思路。

基于谱聚类的主动配电网多时间尺度无功优化策略

高比例分布式光伏接入配电网后,传统优化方案无法有效平抑电压波动,分布式光伏逆变器的无功调控能力难以充分利用。为此,提出一种基于谱聚类的主动配电网多时间尺度无功优化策略,该方法分为日前优化和日内实时优化两个阶段。首先,对离散设备的时间耦合性进行解耦,以配电网网损、平均电压偏差、电压波动严重程度为目标函数,建立基于社交网络搜索算法的日前无功优化模型,确定离散设备静态最优档位序列;其次,通过谱聚类的方法进行耦合,确定离散设备动态最优档位序列,结合改进的分布式光伏逆变器就地控制策略,建立日内实时优化模型,从而抑制日前预测数据偏差导致的电压波动;最后,基于改进后的IEEE33节点系统进行仿真实验。仿真结果表明,所提策略可以有效降低运算难度、提高求解效率,验证了该策略的有效性和优越性。

基于改进VMD-MCKD和深度残差网络的风机齿轮箱故障诊断

行星齿轮箱是风电机组传动系统中的重要部件,其运行工况复杂,背景噪声大,导致齿轮早期故障信号微弱且极易受背景噪声的影响。针对风电机组齿轮箱早期故障特征难以有效提取,齿轮故障难以识别的问题,提出一种风机齿轮箱故障诊断方法。首先,通过变分模态分解算法(variational mode decomposition,VMD)分解风机齿轮箱原始振动信号,获得振动信号故障的最优模态分量;接着,利用最大相关峭度解卷积算法(maximum correlated kurtosis decnvolution,MCKD)通过解卷积重构最优模态分量,削弱背景噪声增强故障冲击成分,获得故障特征;同时利用麻雀搜索算法(sparrow search algorithm,SSA)优化惩罚因子α、模态分解个数K、滤波器阶数L和反褶积周期T等参数,提升振动信号故障特征提取的准确度;最后,构建基于深度残差网络(deep residual network,ResNet)的齿轮箱故障诊断模型,建立齿轮箱故障特征与类别的非线性映射关系,实现风机齿轮箱故障分类识别。实验结果表明,所提风机齿轮箱故障诊断方法的准确率达到97.48%,相较其他方法在信号特征提取和故障诊断效率方面有明显提高。

基于DBN和BES-LSSVM的矿用压风机异常状态识别方法

针对矿用压风机这类分布式系统的异常类别复杂、识别精度低等问题,提出了一种基于深度置信网络(DBN)和最小二乘支持向量机(LSSVM)的异常状态识别方法。首先,分析压风机组成系统及其运行机理,确定常见的异常状态类型;其次,采用DBN无监督学习方式充分挖掘监测数据中异常特征并快速提取;然后,利用秃鹰搜索算法(BES)优化LSSVM的超参数,构建最优的BES-LSSVM分类模型;最后,将DBN提取的异常特征作为BES-LSSVM模型的输入,对矿用压风机异常状态进行识别。试验验证与对比分析结果表明,相较于GA,PSO,GWO算法,BES算法的求解精度和收敛速度均有所提高,同时DBN-BES-LSSVM模型在测试集上平均识别精度达到94.65%,较PCA-LSSVM模型、DBN模型和DBN-LSSVM模型的识别精度分别提高了10.53%,5.84%和3.76%,验证了DBN-BES-LSSVM模型在矿用压风机异常特征提取以及特征识别方面的优越性。

基于端到端深度神经网络和图搜索的OCT图像视网膜层边界分割方法

视网膜层边界的形态变化是眼部视网膜疾病出现的重要标志,光学相干断层扫描(optical coherence tomography,OCT)图像可以捕捉其细微变化,基于OCT图像的视网膜层边界分割能够辅助相关疾病的临床判断.在OCT图像中,由于视网膜层边界的形态变化多样,其中与边界相关的关键信息如上下文信息和显著性边界信息等对层边界的判断和分割至关重要.然而已有分割方法缺乏对以上信息的考虑,导致边界不完整和不连续.针对以上问题,提出一种“由粗到细”的基于端到端深度神经网络和图搜索(graph search,GS)的OCT图像视网膜层边界分割方法,避免了非端到端方法中普遍存在的“断层”现象.在粗分割阶段,提出一种端到端的深度神经网络—注意力全局残差网络(attention global residual network,AGR-Net),以更充分和有效的方式提取上述关键信息.具体地,首先设计一个全局特征模块(global feature module,GFM),通过从图像的4个方向扫描以捕获OCT图像的全局上下文信息;其次,进一步将通道注意力模块(channel attention module,CAM)与全局特征模块串行组合并嵌入到主干网络中,以实现视网膜层及其边界的上下文信息的显著性建模,有效解决OCT图像中由于视网膜层形变和信息提取不充分所导致的误分割问题.在细分割阶段,采用图搜索算法去除AGR-Net粗分割结果中的孤立区域或和孔洞等,保持边界的固定拓扑结构和连续平滑,以实现整体分割结果的进一步优化,为医学临床的诊断提供更完整的参考.最后,在两个公开数据集上从不同的角度对所提出的方法进行性能评估,并与最新方法进行比较.对比实验结果也表明所提方法在分割精度和稳定性方面均优于现有方法.

基于“车-路-站-网”信息耦合的电动汽车有序充电策略

为了消解规模化电动汽车(electrical vehicle,EV)无序充电对交通路网、充电站和配电网运行稳定性带来的负面影响,提出一种基于“车-路-站-网”信息耦合的电动汽车有序充电策略。首先,构建“车-路-站-网”信息耦合模型和动态Floyd最短时间路径搜索模型,为EV用户搜寻最短耗时路径。其次,基于“车-路-站-网”实时状态预测EV用户选择不同路径前往各充电站快充产生的充电决策因素,通过层次分析法和改进CRITIC法综合EV用户充电决策因素的主客观权重,利用Topsis方法决策EV用户的最优充电路径。最后,提出EV用户慢充优化策略,对返程EV用户的慢充负荷进行优化,结合EV慢充和快充负荷,进一步实现配电网负荷的削峰填谷。仿真结果表明,所提出的EV有序充电策略能够同时提升“车-路-站-网”多方运行水平。

基于改进松鼠搜索算法优化神经网络的数控机床进给系统热误差预测

为探究数控机床进给系统中各因素对热误差的影响规律,建立精准的热误差预测模型。在进给速度为10 m/min、环境温度20℃的条件下进行进给系统热误差测量实验,获得进给系统关键点的温升及热误差。为提高预测精度,采用Tent混沌改进松鼠搜索算法,并利用改进的算法对神经网络进行优化,建立热误差预测模型。利用热误差测量实验获得的数据进行验证,结果表明改进前的神经网络预测误差为12.23%,改进后的模型预测误差为8.92%,精度有较大提升。利用预测模型针对不同进给速度下相同位置处热误差进行分析,结果表明,进给系统中关键测温点的温度和丝杠各点的热误差随着进给速度的增加而增加。因此提出的预测模型可实现进给系统热误差的准确预测,为误差补偿提供理论依据。

基于有偏采样的连续进化神经架构搜索

由于需要对每一个搜索到的架构进行独立的性能评估,神经架构搜索(NAS)往往需要耗费大量的时间和计算资源。提出一种基于有偏采样的连续进化NAS方法(OEvNAS)。OEvNAS在架构搜索过程中维护一个超网络,搜索空间中所有的神经网络架构都是该超网络的子网络。在演化计算的每一代对超网络进行少量的训练,子网络直接继承超网络的权重进行性能评估而无需重新训练。为提高超网络的预测性能,提出一种基于有偏采样的超网络训练策略,以更大的概率训练表现优异的网络,在减少权重耦合的同时提高训练效率。此外,设计一种新颖的交叉变异策略来提高算法的全局探索能力。在NATS-Bench和可微分架构搜索(DARTS)两个搜索空间上验证OEvNAS的性能。实验结果表明,OEvNAS的性能超越了对比的主流算法。在NATS-Bench搜索空间上,提出的超网络训练策略在CIFAR-10、CIFAR-100和ImageNet16-200上均取得了优异的预测性能;在DARTS搜索空间上,搜索到的最优神经网络架构在CIFAR-10和CIFAR-100上分别取得了97.67%和83.79%的分类精度。