MSADCN:Multi-Scale Attentional Densely Connected Network for Automated Bone Age Assessment

Bone age assessment(BAA)helps doctors determine how a child’s bones grow and develop in clinical medicine.Traditional BAA methods rely on clinician expertise,leading to time-consuming predictions and inaccurate results.Most deep learning-based BAA methods feed the extracted critical points of images into the network by providing additional annotations.This operation is costly and subjective.To address these problems,we propose a multi-scale attentional densely connected network(MSADCN)in this paper.MSADCN constructs a multi-scale dense connectivity mechanism,which can avoid overfitting,obtain the local features effectively and prevent gradient vanishing even in limited training data.First,MSADCN designs multi-scale structures in the densely connected network to extract fine-grained features at different scales.Then,coordinate attention is embedded to focus on critical features and automatically locate the regions of interest(ROI)without additional annotation.In addition,to improve the model’s generalization,transfer learning is applied to train the proposed MSADCN on the public dataset IMDB-WIKI,and the obtained pre-trained weights are loaded onto the Radiological Society of North America(RSNA)dataset.Finally,label distribution learning(LDL)and expectation regression techniques are introduced into our model to exploit the correlation between hand bone images of different ages,which can obtain stable age estimates.Extensive experiments confirm that our model can converge more efficiently and obtain a mean absolute error(MAE)of 4.64 months,outperforming some state-of-the-art BAA methods.

Underwater Image Enhancement Based on Multi-scale Adversarial Network

In this study,an underwater image enhancement method based on multi-scale adversarial network was proposed to solve the problem of detail blur and color distortion in underwater images.Firstly,the local features of each layer were enhanced into the global features by the proposed residual dense block,which ensured that the generated images retain more details.Secondly,a multi-scale structure was adopted to extract multi-scale semantic features of the original images.Finally,the features obtained from the dual channels were fused by an adaptive fusion module to further optimize the features.The discriminant network adopted the structure of the Markov discriminator.In addition,by constructing mean square error,structural similarity,and perceived color loss function,the generated image is consistent with the reference image in structure,color,and content.The experimental results showed that the enhanced underwater image deblurring effect of the proposed algorithm was good and the problem of underwater image color bias was effectively improved.In both subjective and objective evaluation indexes,the experimental results of the proposed algorithm are better than those of the comparison algorithm.

Novel Path Counting-Based Method for Fractal Dimension Estimation of the Ultra-Dense Networks

Next-generation networks,including the Internet of Things(IoT),fifth-generation cellular systems(5G),and sixth-generation cellular systems(6G),suf-fer from the dramatic increase of the number of deployed devices.This puts high constraints and challenges on the design of such networks.Structural changing of the network is one of such challenges that affect the network performance,includ-ing the required quality of service(QoS).The fractal dimension(FD)is consid-ered one of the main indicators used to represent the structure of the communication network.To this end,this work analyzes the FD of the network and its use for telecommunication networks investigation and planning.The clus-ter growing method for assessing the FD is introduced and analyzed.The article proposes a novel method for estimating the FD of a communication network,based on assessing the network’s connectivity,by searching for the shortest routes.Unlike the cluster growing method,the proposed method does not require multiple iterations,which reduces the number of calculations,and increases the stability of the results obtained.Thus,the proposed method requires less compu-tational cost than the cluster growing method and achieves higher stability.The method is quite simple to implement and can be used in the tasks of research and planning of modern and promising communication networks.The developed method is evaluated for two different network structures and compared with the cluster growing method.Results validate the developed method.

结合强化学习和DenseNet的远程监督关系抽取模型

关系抽取是信息获取领域的重要任务之一。为了更好地解决数据集中的噪声问题和句子深层次语义表征,提出一种结合强化学习和密集连接卷积神经网络的远程监督关系抽取模型,模型分为句子选择器和关系分类器。在句子选择器中,基于强化学习的方法能有效过滤噪声语句,提升输入数据质量;在关系分类器中,通过DenseNet深层网络中的特征复用,学习更丰富的语义特征。在NYT数据集上的实验结果表明句子选择器能够有效过滤噪声,该模型的关系抽取性能相比基线模型得到有效提高。

Multi-Scale Attention-Based Deep Neural Network for Brain Disease Diagnosis

Whole brain functional connectivity(FC)patterns obtained from resting-state functional magnetic resonance imaging(rs-fMRI)have been widely used in the diagnosis of brain disorders such as autism spectrum disorder(ASD).Recently,an increasing number of studies have focused on employing deep learning techniques to analyze FC patterns for brain disease classification.However,the high dimensionality of the FC features and the interpretation of deep learning results are issues that need to be addressed in the FC-based brain disease classification.In this paper,we proposed a multi-scale attention-based deep neural network(MSA-DNN)model to classify FC patterns for the ASD diagnosis.The model was implemented by adding a flexible multi-scale attention(MSA)module to the auto-encoder based backbone DNN,which can extract multi-scale features of the FC patterns and change the level of attention for different FCs by continuous learning.Our model will reinforce the weights of important FC features while suppress the unimportant FCs to ensure the sparsity of the model weights and enhance the model interpretability.We performed systematic experiments on the large multi-sites ASD dataset with both ten-fold and leaveone-site-out cross-validations.Results showed that our model outperformed classical methods in brain disease classification and revealed robust intersite prediction performance.We also localized important FC features and brain regions associated with ASD classification.Overall,our study further promotes the biomarker detection and computer-aided classification for ASD diagnosis,and the proposed MSA module is flexible and easy to implement in other classification networks.

基于DenseNet与PointNet融合算法的三维点云分割

点云分割对于智能驾驶、物体检测和识别、逆向工程等任务非常重要。PointNet是一种能够直接处理点云数据的方法,近年来在点云分割任务中得到广泛应用,但其分割精度较低,而PointNet++的计算成本又较高。针对以上问题,提出一种融合DenseNet和PointNet的算法,用于点云分割,并引入三分支混合注意力机制,以提高PointNet在提取局部特征方面的能力。基于密集连接卷积网络(DenseNet)思想,提出用DenseNet-STN和DenseNet-MLP结构来替代PointNet中的空间变换网络(STN)和多层感知机(MLP);同时,使用Add连接代替密集块(DenseBlock)中的Concat连接,以提高对点特征间相关性的准确性,同时不显著增加模型复杂度。DenseNet-PointNet能够提高复杂分类问题的泛化能力,实现对复杂函数更好的逼近,从而提高点云分割的准确率。有效性和消融实验结果表明,本文算法具有良好的性能。点云分割实验结果表明,DenseNet-PointNet在大多数类别中的交并比(IoU)都高于PointNet的IoU,并在部分类别中也高于PointNet++,参数量是PointNet++的47.6%,浮点运算量(FLOPs)是PointNet++的49.1%。实验结果验证了DenseNet-PointNet的可行性和有效性。

基于DenseNet卷积神经网络的短期风电预测方法

风能作为一种清洁、可再生的能源,在能源转型中扮演着至关重要的角色,准确预测风电出力对电力系统的安全高效运行非常重要,然而风速的波动性和随机性,对风电预测带来了挑战。为了提高风电预测的准确性,提出了一种基于DenseNet卷积神经网络的短期风电预测模型。该模型通过精简DenseNet201网络得到了拥有出色的密集连接结构和适当深度、宽度的DenseNet160网络,不仅能缓解训练过程中梯度消失现象,还能通过密集连接将浅层的信息反映到深层,实现深度监督。基于巴西纳塔尔地区378 d的风力数据集,采用DenseNet160网络以及27种算法对未来一天的风力发电情况进行预测。结果表明:DenseNet160网络的平均绝对误差、均方误差以及平均绝对百分误差比其他算法分别降低了至少10.89%,4.98%,8.68%;同时,与使用相同数据集的混合经济模型相比,DenseNet160网络的MAE值小了25.56%。说明该模型能精准地拟合风力发电数据,获得可靠的风力预测结果。

基于DSC-DenseNet的流程工业系统故障监测

田纳西-伊士曼过程数据高纬度、高耦合,存在数据特征难以提取的问题。为进一步提高流程工业系统中故障监测的识别率,现将一维稠密卷积网络(1D-DenseNet)与深度可分离卷积(DSC)结合,利用DenseNet的高效特征提取能力,并结合DSC减少计算参数、提高诊断效率,以提供基于DSC-DenseNet的故障监测方式。先将数据进行归一化整理,并加入随机种子避免过拟合,随后将处理后的结果作为DSC-DenseNet的输入进行特征提取,然后将输出结果传入全连接层进行故障分类;最后在TEP数据集上进行准确率测试。结果证明:基于DSC-DenseNet的方法能有效分辨故障类型,故障分类准确率达到98.8%。并证明DSC-DenseNet比传统DenseNet有更好的故障识别效果。

基于多任务NR-DenseNet网络的航班延误预测模型

不同于目前大多数只倾向于研究单一的分类或回归任务的航班延误预测方法,该文提出一种基于多任务NR-DenseNet网络的航班延误预测模型,旨在同时实现航班延误等级分类预测与延误时间回归预测。首先,预处理相关数据;其次,建立多任务学习特征提取共享层,使用NR-DenseNet网络提取任务之间的共享参数,深度挖掘任务之间的相关特征;然后,建立多任务学习特定任务层,通过回归器与分类器分别输出特定任务的预测结果;最后,采用损失加权方法对两个任务损失函数进行优化,平衡任务间的收敛速度,提高模型泛化性。将模型应用在宁波机场数据集中,与单任务模型相比回归任务平均MSE降低了23.4%,平均MAE降低了14.2%,分类平均准确率提升了2.7%。实验结果表明,该文方法提升了分类任务的准确率降低了回归任务的误差,可以有效提升模型性能。

DenseNet结合空间通道注意力机制的环境声音分类

音乐信息识别(MIR)和自动语音识别(ASR)都是以结构化声音为特点的声音识别,环境声音识别在声音识别领域的难度很大。为了充分利用从环境声中提取的Log-Mel谱图的空间特征与通道特征,提出了一种基于密集连接卷积网络(DenseNet)的空间通道注意力机制。使用DenseNet对Log-Mel谱图进行特征提取,引入空间通道注意力机制使网络更加关注显著特征;为了解决数据不足导致的过拟合问题,将混合数据增强的方法应用于Log-Mel谱图,从而保证了数据的多样性;在2个公共数据集(ESC-50和ESC-10)验证所提方法的有效性。结果表明:所提的空间通道注意力机制模型能够使神经网络对环境声音的识别率分别达到79.3%(ESC-50)和94.3%(ESC-10)。

基于DenseNet与声学层析成像的温度场高分辨率重建

提出一种虚拟观测结合密集连接网络(DenseNet)的两阶段声学层析成像温度场高分辨率重建算法。以总体最小二乘为目标,采用虚拟观测法对超声飞行时间(TOF)进行重建,得到粗网格下的温度分布;然后利用搭建的DenseNet预测细化网格的温度分布信息,采用双输入模型,同时利用模块化输出方法对典型的温度场模型进行数值仿真。结果表明:该算法的重建质量和抗噪性均优于常用的虚拟观测方法、Landweber迭代法、Tikhonov算法及代数重建(ART)方法。

基于CVMD-GRU-DenseNet混合模型的短期电力负荷预测

针对电力负荷时序变化非线性和多周期性特点,提出一种基于分解-预测-重构框架的CVMD-GRU-DenseNet短期负荷预测方法。分解阶段依据子序列间的相关熵确定VMD最佳分解数,提高负荷序列分解质量;预测阶段针对各子序列特点筛选输入特征,规律性强的低频分量采用GRU神经网络预测模型,强随机性的高频分量采用DenseNet神经网络预测模型;最后将各分量的预测结果重构为负荷预测曲线。湖北某市四季的实际负荷算例结果表明,该方法能有效提高短期负荷预测精度,并具有较强的泛化能力。

基于SRDenseNet的指纹超分辨率重建研究

指纹识别是一种常用的生物特征识别技术,图像质量对指纹识别效果有着重要的影响。针对现有指纹图像质量低导致识别准确率差的问题,提出一种基于SRDenseNet的指纹超分辨率重建算法。该方法使用深度卷积神经网络为基本结构,加入了Dense块和Dense跳连接以充分利用多层次的特征信息,提出将全局均方差和平均局部结构相似度引入损失函数,有效提升深度卷积神经网络对指纹图像的分辨率提高与全局结构和局部细节重建的能力。在FVC数据集上进行测试和验证,将该方法与现有其他方法进行比较。结果表明,2、3、4倍率的超分辨率指纹图像的匹配等错误率分别从原来的5.456%、8.730%和16.091%下降至4.762%、7.500%和12.540%,实验证明了该方法的有效性。

Detection of influential nodes with multi-scale information

The identification of influential nodes in complex networks is one of the most exciting topics in network science.The latest work successfully compares each node using local connectivity and weak tie theory from a new perspective.We study the structural properties of networks in depth and extend this successful node evaluation from single-scale to multi-scale.In particular,one novel position parameter based on node transmission efficiency is proposed,which mainly depends on the shortest distances from target nodes to high-degree nodes.In this regard,the novel multi-scale information importance(MSII)method is proposed to better identify the crucial nodes by combining the network's local connectivity and global position information.In simulation comparisons,five state-of-the-art algorithms,i.e.the neighbor nodes degree algorithm(NND),betweenness centrality,closeness centrality,Katz centrality and the k-shell decomposition method,are selected to compare with our MSII.The results demonstrate that our method obtains superior performance in terms of robustness and spreading propagation for both real-world and artificial networks.

基于LW-DenseNet的采煤机摇臂齿轮故障诊断

为了提升采煤机摇臂齿轮故障诊断准确率、减小模型尺寸且方便部署到更多移动端与边缘设备上,搭建了基于轻量化密集连接卷积网络(LW-DenseNet)的采煤机摇臂齿轮故障诊断模型。采用可分离卷积代替传统卷积减少模型参数,提高诊断效率;通过密集连接机制增强特征传播,加强特征提取能力。利用采煤机摇臂加载试验台采集的摇臂齿轮振动信号进行训练并验证模型的有效性。实验结果表明,与多种诊断模型比较,所提方法仅以0.05 MB的模型大小即可达到99.276%的分类精度,并利用凯斯西储大学轴承数据集验证了模型具有良好的泛化性。最后对关键层利用t-SNE进行可视化表示,清晰地展现了模型良好的特征提取性能。

Attention-based neural network for end-to-end music separation

The end-to-end separation algorithm with superior performance in the field of speech separation has not been effectively used in music separation.Moreover,since music signals are often dual channel data with a high sampling rate,how to model longsequence data and make rational use of the relevant information between channels is also an urgent problem to be solved.In order to solve the above problems,the performance of the end-to-end music separation algorithm is enhanced by improving the network structure.Our main contributions include the following:(1)A more reasonable densely connected U-Net is designed to capture the long-term characteristics of music,such as main melody,tone and so on.(2)On this basis,the multi-head attention and dualpath transformer are introduced in the separation module.Channel attention units are applied recursively on the feature map of each layer of the network,enabling the network to perform long-sequence separation.Experimental results show that after the introduction of the channel attention,the performance of the proposed algorithm has a stable improvement compared with the baseline system.On the MUSDB18 dataset,the average score of the separated audio exceeds that of the current best-performing music separation algorithm based on the time-frequency domain(T-F domain).

基于UNet3+生成对抗网络的视频异常检测

为解决传统视频异常检测方法在不同场景下多尺度特征提取不完全的问题,提出两种方法:一种是用于简单场景的基于UNet3+的生成对抗网络方法(简称U3P^(2)),另一种是用于复杂场景的基于UNet++的生成对抗网络方法(简称UP^(3))。两种方法分别对连续输入的视频帧生成预测,引入多种损失函数和光流模型学习其外观与运动信息,通过计算AUC进行性能评估。U3P^(2)方法以6.3 M参数量在Ped2数据集的AUC提升约0.6%,而UP^(3)方法在Avenue数据集的AUC提升约0.8%,验证其能够有效应对不同场景下的异常检测任务。

结合句法增强与图注意力网络的方面级情感分类

方面级情感分类旨在识别给定特定方面文本的情感极性,在本领域中,将图神经网络与句法依赖解析相结合是当下热门的研究方向之一,此类方法通过句法解析捕捉句子中词与词之间的关系,依此构建图结构,输入图神经网络中得到情感极性。若句法解析器出现解析错误,将会对以图为基础的图神经网络模型产生巨大影响。为了增强解析器生成的句法依赖树的解析结果,文中提出了一种句法增强图注意力网络,该网络通过融合多个解析器的解析结果,提高句法依赖解析精度,得到更精准的依赖关系句法图;在图注意力网络中使用密集连接机制捕获更丰富的特征,更适配于增强后的句法图,同时引入方面注意力机制捕获方面语义特征。实验结果验证了句法增强方法的有效性,在3个基准数据集上的分类准确度都有所提高,在方面级情感分析领域具有较好的表现。

金字塔方差池化网络的图像超分辨率重建

为减少高频信息丢失对图像重建造成的影响,进一步增强对特征信息的挖掘,以金字塔方差池化模块为核心构建了一个生成网络。首先,该网络利用不同级别的方差池化挖掘低分辨率图像蕴含的特征信息,并结合金字塔结构获取不同尺度与不同子区域的上下文信息,从而进一步丰富特征信息量;然后,利用密集连接结构增强特征信息之间的关联性,以提高网络的表达能力;最后,引入组归一化操作来加强网络的收敛性。实验结果表明,该模型与其他方法在Set5、Set14、DIV2K100公开测试集上进行比较,在放大倍数因子为4时,峰值信噪比平均提高了0.509 dB,结构相似性平均提高了0.016。所提模型不仅在峰值信噪比和结构相似性上有一定的提高,其重建图像在视觉效果上也拥有更多的真实细节。

基于多尺度密集连接网络的电容层析成像图像重建

为求解电容层析成像中的非线性病态反问题,提出了一种多尺度密集连接网络(MD-Net)模型。该模型由多尺度特征融合模块和密集连接块组成,以通过融合多尺度特征进一步提高图像的重建精度。通过MATLAB仿真实验平台构建了流型数据集,利用密集连接网络的非线性映射能力,完成训练集的学习与训练,并利用测试集进行训练效果评价。在此基础上进行了静态实验。仿真与静态实验结果均表明:与LBP、Landweber迭代算法和其他深度学习方法相比,该方法的重建精度最高、抗噪能力强,并具有良好的泛化能力。