Evolution and Effectiveness of Loss Functions in Generative Adversarial Networks

Recently,the evolution of Generative Adversarial Networks(GANs)has embarked on a journey of revolutionizing the field of artificial and computational intelligence.To improve the generating ability of GANs,various loss functions are introduced to measure the degree of similarity between the samples generated by the generator and the real data samples,and the effectiveness of the loss functions in improving the generating ability of GANs.In this paper,we present a detailed survey for the loss functions used in GANs,and provide a critical analysis on the pros and cons of these loss functions.First,the basic theory of GANs along with the training mechanism are introduced.Then,the most commonly used loss functions in GANs are introduced and analyzed.Third,the experimental analyses and comparison of these loss functions are presented in different GAN architectures.Finally,several suggestions on choosing suitable loss functions for image synthesis tasks are given.

Brain Functional Network Generation Using Distribution-Regularized Adversarial Graph Autoencoder with Transformer for Dementia Diagnosis

The topological connectivity information derived from the brain functional network can bring new insights for diagnosing and analyzing dementia disorders.The brain functional network is suitable to bridge the correlation between abnormal connectivities and dementia disorders.However,it is challenging to access considerable amounts of brain functional network data,which hinders the widespread application of data-driven models in dementia diagnosis.In this study,a novel distribution-regularized adversarial graph auto-Encoder(DAGAE)with transformer is proposed to generate new fake brain functional networks to augment the brain functional network dataset,improving the dementia diagnosis accuracy of data-driven models.Specifically,the label distribution is estimated to regularize the latent space learned by the graph encoder,which canmake the learning process stable and the learned representation robust.Also,the transformer generator is devised to map the node representations into node-to-node connections by exploring the long-term dependence of highly-correlated distant brain regions.The typical topological properties and discriminative features can be preserved entirely.Furthermore,the generated brain functional networks improve the prediction performance using different classifiers,which can be applied to analyze other cognitive diseases.Attempts on the Alzheimer’s Disease Neuroimaging Initiative(ADNI)dataset demonstrate that the proposed model can generate good brain functional networks.The classification results show adding generated data can achieve the best accuracy value of 85.33%,sensitivity value of 84.00%,specificity value of 86.67%.The proposed model also achieves superior performance compared with other related augmentedmodels.Overall,the proposedmodel effectively improves cognitive disease diagnosis by generating diverse brain functional networks.

Generative Adversarial Network-Based Electromagnetic Signal Classification: A Semi- Supervised Learning Framework

Generative adversarial network(GAN)has achieved great success in many fields such as computer vision,speech processing,and natural language processing,because of its powerful capabilities for generating realistic samples.In this paper,we introduce GAN into the field of electromagnetic signal classification(ESC).ESC plays an important role in both military and civilian domains.However,in many specific scenarios,we can’t obtain enough labeled data,which cause failure of deep learning methods because they are easy to fall into over-fitting.Fortunately,semi-supervised learning(SSL)can leverage the large amount of unlabeled data to enhance the classification performance of classifiers,especially in scenarios with limited amount of labeled data.We present an SSL framework by incorporating GAN,which can directly process the raw in-phase and quadrature(IQ)signal data.According to the characteristics of the electromagnetic signal,we propose a weighted loss function,leading to an effective classifier to realize the end-to-end classification of the electromagnetic signal.We validate the proposed method on both public RML2016.04c dataset and real-world Aircraft Communications Addressing and Reporting System(ACARS)signal dataset.Extensive experimental results show that the proposed framework obtains a significant increase in classification accuracy compared with the state-of-the-art studies.

Estimating Weibull Parameters Using Least Squares and Multilayer Perceptron vs. Bayes Estimation

The Weibull distribution is regarded as among the finest in the family of failure distributions.One of the most commonly used parameters of the Weibull distribution(WD)is the ordinary least squares(OLS)technique,which is useful in reliability and lifetime modeling.In this study,we propose an approach based on the ordinary least squares and the multilayer perceptron(MLP)neural network called the OLSMLP that is based on the resilience of the OLS method.The MLP solves the problem of heteroscedasticity that distorts the estimation of the parameters of the WD due to the presence of outliers,and eases the difficulty of determining weights in case of the weighted least square(WLS).Another method is proposed by incorporating a weight into the general entropy(GE)loss function to estimate the parameters of the WD to obtain a modified loss function(WGE).Furthermore,a Monte Carlo simulation is performed to examine the performance of the proposed OLSMLP method in comparison with approximate Bayesian estimation(BLWGE)by using a weighted GE loss function.The results of the simulation showed that the two proposed methods produced good estimates even for small sample sizes.In addition,the techniques proposed here are typically the preferred options when estimating parameters compared with other available methods,in terms of the mean squared error and requirements related to time.

Image Rain Removal Using Conditional Generative Networks Incorporating

The research of removing rain from pictures or videos has always been an important topic in the field of computer vision and image processing. Most noise reduction methods more or less remove texture details in rain-free areas, resulting in an over-smoothing effect in the restored background. The research on image noise removal is very meaningful. We exploit the powerful generative power of a modified generative adversarial network (CGAN) by enforcing an additional condition that makes the derained image indistinguishable from its corresponding ground-truth clean image. An efficient and lightweight attention machine mechanism NAM is introduced in the generator, and an IDN-CGAN model is proposed to capture image salient features through attention operations. Taking advantage of the mutual information in different dimensions of the features to further suppress insignificant channels or pixels to ensure better visual quality, we also introduce a new fine-grained loss function in the generator-discriminator pair, predicting and real data degree of disparity to achieve improved results.

应用生成对抗网络的地震数据重建和去噪一体化方法

在实际采集过程中,受地形条件和人为因素的影响,地震数据不仅在空间上会出现采样不足或不规则的情况,而且会混入噪声,不利于后续地震数据的处理和解释。通常将重建与去噪分为两个阶段处理,这样往往会引入额外的误差。为此,文中提出了一种基于条件韦氏生成对抗网络(cWGAN)的地震数据重建去噪一体化方法,该方法研究的重点是在缺失道和噪声的混合干扰下,准确提取地震数据的有效特征。首先,以U-Net模型为基本网络结构来构建生成器模型,分级提取地震数据同相轴特征;在判别器模型中引入条件约束,引导生成器优化梯度方向。其次,建立重建和去噪误差描述模型,该模型设计了一体化损失函数,可以兼顾重建与去噪两方面的处理任务。最后,经过合成数据和实际数据测试,证明文中所提的网络模型恢复的地震数据信噪比更高且具有较强鲁棒性。

融合IMR-WGAN的时序数据修复方法

工业数据由于技术故障和人为因素通常导致数据异常,现有基于约束的方法因约束阈值设置的过于宽松或严格会导致修复错误,基于统计的方法因平滑修复机制导致对时间步长较远的异常值修复准确度较低.针对上述问题,提出了基于奖励机制的最小迭代修复和改进WGAN混合模型的时序数据修复方法.首先,在预处理阶段,保留异常数据,进行信息标注等处理,从而充分挖掘异常值与真实值之间的特征约束.其次,在噪声模块提出了近邻参数裁剪规则,用于修正最小迭代修复公式生成的噪声向量.将其传递至模拟分布模块的生成器中,同时设计了一个动态时间注意力网络层,用于提取时序特征权重并与门控循环单元串联组合捕捉不同步长的特征依赖,并引入递归多步预测原理共同提升模型的表达能力;在判别器中设计了Abnormal and Truth奖励机制和Weighted Mean Square Error损失函数共同反向优化生成器修复数据的细节和质量.最后,在公开数据集和真实数据集上的实验结果表明,该方法的修复准确度与模型稳定性显著优于现有方法.

基于字符和词特征融合的恶意域名检测

针对现有恶意域名检测方法对域名生成算法(domain generation algorithm, DGA)随机产生的恶意域名检测性能不高,且对由随机单词组成的恶意域名检测效果较差的问题,提出一种基于字符和词特征融合的恶意域名检测算法(cha-racter and word network, CWNet)。利用并行卷积神经网络分别提取域名中字符和词的特征;将两种特征进行拼接,构造成融合特征;利用Softmax函数实现合法域名与恶意域名的检测。实验结果表明,该算法可以提升对恶意域名的检测能力,对更具挑战性的恶意域名家族的检测准确率提升效果更为明显。

基于生成对抗网络的人脸妆容迁移方法研究

妆容迁移是一项利用计算机视觉和深度学习算法将一种妆容的风格转移到其他人脸上的技术,以实现高仿妆效果转换。为了有效解决现有人脸妆容迁移方法中存在的上妆区域错误和妆容迁移不完整的问题,提出了一种基于生成对抗网络的人脸妆容迁移方法(MutNet)。以解决妆容迁移不完整的问题为目标,该方法在解码器中引入了空间注意力机制,来帮助网络更加聚焦于需要修改的区域,并通过引入孪生对比损失,更好地实现人脸之间的语义对应关系,有效缓解或克服上妆区域错误的问题。同时通过与其他方法的对比结果表明,MutNet能获得更协调的上妆效果。

基于CSLS-CycleGAN的侧扫声纳水下目标图像样本扩增法

针对侧扫声纳水下目标图像稀缺,获取难度大、成本高,导致基于深度学习的目标检测模型性能差的问题,结合光学域类目标数据集丰富的现状,提出一种基于通道和空间注意力(channel and spatial attention,CSA)模块、最小二乘生成对抗生成网络(least squares generative adversarial networks,LSGAN)及循环对抗生成网络(cycle generative adversarial networks,CycleGAN)的侧扫声纳水下目标图像样本扩增方法。首先,受CycleGAN的启发,设计基于循环一致性的单循环网络结构,保证模型的训练效率。然后,在生成器中融合CSA模块,减少信息弥散的同时增强跨纬度交互。最后,设计了基于LSGAN的损失函数,提高生成图像质量的同时提高训练稳定性。在船舶光学域数据集与侧扫声纳沉船数据集上进行实验,所提方法实现了光学-侧扫声纳样本间信息的高效、稳健转换以及大量侧扫声纳目标样本的扩增。同时,基于本文生成样本训练后的检测模型进行了水下目标检测,结果表明,使用本文样本扩增数据训练后的模型在少样本沉船目标检测的平均准确率达到了84.71%,证明了所提方法实现了零样本和小样本水下强代表性目标样本的高质量扩增,并为高性能水下目标检测模型构建提供了一种新的途径。

基于生成对抗网络的非织造布二维图像重建

非织造布纤维结构的准确表征是其性能分析的重要基础。为了解决基于近似模拟的表征结果中纤维形态、结构与真实样本不一致的问题,提出了一种基于生成对抗网络的非织造布二维图像重建方法。使用全自动光学显微镜对非织造布图像进行抓取,并在此基础上构建纤维生成对抗网络(Fiber generation adversarial network,FGAN)对图像样本进行建模。针对高分辨率图像重建时存在的失真问题,采用多尺度训练策略,同时引入权重多样性损失。采用图像质量评估指标FID作为实验评价指标,分别与DCGAN、WGAN-GP、BEGAN、PROGAN等生成模型进行对比实验。结果表明:FGAN重建的非织造布图像质量更高;消融实验证明,多尺度训练策略与权重多样性损失函数中FID数值分别降低24.52%、20.31%。FGAN模型的提出,使非织造布结构分析摆脱对近似模拟方法的依赖,提供了准确的纤维分布信息,对非织造布的质量评估、性能优化等应用具有重要意义。

基于生成对抗网络的高精度室内无线定位方法

虚拟标签、参考标签布局形式不一致是导致网络主机难以准确定位室内空间节点的主要原因,针对这一问题,完善生成对抗网络的结构体系,并在求解相应损失函数的基础上,根据虚拟标签、参考标签的布局形式,准确定位空间节点,实现基于生成对抗网络的室内空间规划。在空间规划后,对室内空间数据实施扩充处理,根据空间数据无线传输精度的数值水平,确定定位阈值的取值范围。通过实验发现,该方法可以保证虚拟标签、参考标签的一致性布局,实现网络主机对室内环境空间节点的准确定位。

科技文献的多层次结构功能识别

实现科技文献结构功能的自动识别有助于提升细粒度信息检索、关键词抽取、引文分析等任务的效率。针对当前结构功能识别研究面临的文本内部依赖关系表达能力较弱、模型泛化迁移能力不足等问题,本研究利用图卷积神经网络捕捉单词节点间存在的固有依赖信息和拓扑结构,提升模型对科技文本建模表达能力,同时,还引入对抗学习思想,提升结构功能识别模型的泛化能力。选取ScienceDirect数据集,考察多种模型方法对章节标题、章节内容、章节段落三个不同层次的结构功能的识别效果,并在PubMed-20k的医学摘要结构功能数据集上进一步测试多种模型的跨领域迁移能力。研究结果表明,在章节标题层次,BERT+GCN的识别效果最佳,F1值达到了88%,比基线模型提升3%;在章节内容层次,BERT+GAN的识别效果最佳,F1值达到了76%,比基线模型提升了3%;在章节段落层次,F1值达到了68%。BERT+GCN的跨领域迁移能力相比其他模型更优,在跨领域数据上取得了90%的F1值。

基于最大均值差异的卷积神经网络故障诊断模型

针对工程场景中轴承故障数据采集困难,小样本下轴承故障诊断准确率较低且稳定性不高的问题,提出了一种小样本下滚动轴承故障的诊断方法,即基于最大均值差异(MMD)的卷积神经网络(CNN)诊断模型(方法)。首先,根据轴承故障机理,获取了滚动轴承故障的仿真信号,基于生成式对抗网络构建了仿真信号与少量真实样本间的对抗训练模型,得到了伪域样本,并将其扩充为训练数据集;其次,以交叉熵损失和最大均值差异(MMD)为卷积神经网络(CNN)的优化准则,引入了缩放因子,对网络进行了动态优化,根据测试结果选取缩放因子为0.05作为最优网络结构参数,构建了故障诊断的训练模型;最后,将结构均为1024个数据点的伪域样本和真实样本共同构成模型的训练集,对其进行了归一化处理,然后将其输入到构建的网络模型中,并以MMD作为约束,进行了卷积、池化操作,以实现特征提取的目的,经反向传播对模型进行了优化,实现了诊断模型参数的迭代更新目标。研究结果表明:基于MMD的CNN诊断模型(方法)对小样本下轴承的故障诊断识别精度有明显的改善,当样本数仅为16时,识别率可达95%以上,证明该方法在小样本下的轴承故障诊断中依然能获得较高的故障识别率。

改进的U-Net网络小断层识别技术在玛湖凹陷玛中地区三叠系白碱滩组的应用

利用改进的U-Net网络小断层识别技术,对准噶尔盆地玛湖凹陷玛中地区三叠系白碱滩组的小断层进行了识别。研究结果表明:(1)构造导向滤波预处理能有效改善地震资料的品质,提高断层识别的准确率。加入了跳跃连接和中继监督、正态标准化和聚焦均方损失函数的U-Net网络方法,对小断层的精细识别能力有所提升。(2)使用200组训练样本集和20组验证样本集,模型地震数据由反射系数与雷克子波褶积生成,断层由人工标注而成。选取最优的网络模型参数,并在合成的含噪地震数据上分别利用相干属性、常规U-Net网络方法及改进的U-Net网络方法进行测试,构造导向滤波有效突出了断层的边界,且增强了同相轴的横向连续性,改进后的U-Net网络方法对于7 m以上断距的断层可进行有效识别。(3)对于玛湖凹陷玛中地区三叠系白碱滩组高角度走滑断裂和伴生小断距次级断裂的识别,改进后的U-Net网络方法的识别精度明显高于相干属性和常规U-Net网络方法,研究区大侏罗沟断裂北翼的(3)号与(4)号砂体,是拓展MZ4井区三叠系白碱滩组高效勘探的有利区。

Skip-cycleGAN:一种果园苹果异源图像配准模型

针对有监督的配准模型的性能受限于给定的标签以及循环一致性生成对抗网络训练不稳定,收敛速度较慢,易过拟合,对复杂场景的图像处理效果不佳的问题,基于循环一致性生成对抗网络从3个方面(生成器、鉴别器和损失函数)进行改进,提出一种无监督的异源图像配准模型。生成网络的下采样与上采样之间引入带有特征转换残差层的跳跃连接,可以确保梯度的有效传递,减少前向与反向传播过程中信息损失,实现低级特征和高级特征的结合,从而缓解梯度消失和梯度爆炸,促进神经网络的收敛,有助于网络学习更多的上下文信息。在一个自建果园苹果数据集和两个公共数据集上对模型进行评估,实验得出在改进后的生成器基础上,对于形变比较大的数据集选取70×70 PatchGAN鉴别器更合适,对于形变比较小的数据集选取PixelGAN鉴别器更合适。与8个经典算法进行对比,用6个性能指标进行评估,实验结果表明该模型在异源果园苹果数据集上的综合表现优于对比算法。未来将提升模型对异源图像亮度和对比度的鲁棒性,并进行轻量化模型的工作。

CI-WGAN:融合临床指标和WGAN的孤独症个体化脑功能连接网络生成

脑功能连接(Functional connectivity,FC)网络作为潜在的脑影像标志物对孤独症谱系障碍(Autism spectrum disorder,ASD)的辅助诊疗研究具有重要作用。然而现有的FC生成方法大多仅基于脑影像数据,未充分考虑个体的临床指标从而易丢失疾病的特异性信息。而且,ASD作为一种谱系障碍,其临床指标存在显著的个体化差异。因此,仅基于脑影像数据的传统生成模型在生成准确的且能反映个体化临床指标的ASD个体FC的任务上存在挑战。针对上述挑战,提出了个体化临床指标引导的沃瑟斯坦生成对抗网络模型(Clinical-indicator-aware Wasserstein generative adversarial network,CI-WGAN),用于生成孤独症个体化FC网络。该模型引入个体化临床指标引导机制,实现了高精度ASD患者FC网络的生成。基于全世界最大孤独症脑影像公开数据集之一的ABIDE I数据集进行实验,CI-WGAN生成FC与真实FC的峰值信噪比(Peak signal-to-noise ratio,PSNR)、结构相似度(Structural similarity,SSIM)与平均绝对误差(Mean absolute error,MAE)分别达到19.037、0.236与0.178,相较于其他FC生成模型分别提升了3%、12%与2%。此外基于生成FC和独立临床验证指标的表征相似度分析(Representational similarity analysis,RSA),CI-WGAN生成的FC相较其他模型生成FC最少提高了0.1倍和3.7倍,证明了CI-WGAN生成的FC包含更多的ASD个体特异性信息。本文提出的CI-WGAN模型实现了高质量个体化FC的生成,为ASD的早期诊断和个性化治疗提供了有力的技术支持。

基于改进生成对抗网络的图像风格迁移算法

图像风格迁移是图像处理领域的研究热点,但目前风格迁移模型存在生成图像细节模糊、风格纹理的色彩效果较差以及模型参数过多等问题。文中提出了一种基于改进循环一致性生成对抗网络的图像风格迁移方法,通过加入Ghost卷积模块和反残差改进模块来优化生成器网络结构,以此降低模型参数量和计算成本。同时能增强网络的特征提取能力,在损失函数中加入内容风格损失项、颜色重建损失项和映射一致性损失项来改善模型的生成能力,提升生成图像质量。实验结果表明,所提改进方法具有较强的风格迁移能力,有效增强了生成图像的内容细节和风格纹理的色彩效果,显著提升了图像质量,模型性能也得到了改善。

基于生成对抗网络的面部图像修复算法研究

为了解决目前基于生成对抗网络的图像修复算法在修复大范围面部图像时修复效果不好,特征提取不充分的问题,提出了基于注意力机制的生成对抗网络修复模型。引入自注意力机制模块来感受图像全局特征用来生成图像缺失区域,能够更好地修复图像的大范围缺失。同时改进了判别器,引入感知损失,提高了修复图像的结构相似度。在CelebA数据集上的实验结果表明,该算法在各项评价指标上均优于现有主流算法,其PSNR损失提高了0.81%~1.94%,SSIM提高了1.06%~2.49%,MSE降低了0.12%~0.35%。

Pedestrian wind flow prediction using spatial-frequency generative adversarial network

Pedestrian wind flow is a critical factor in designing livable residential environments under growing complex urban conditions.Predicting pedestrian wind flow during the early design stages is essential but currently suffers from inefficiencies in numerical simulations.Deep learning,particularly generative adversarial networks(GAN),has been increasingly adopted as an alternative method to provide efficient prediction of pedestrian wind flow.However,existing GAN-based wind flow prediction schemes have limitations due to the lack of considering the spatial and frequency characteristics of wind flow images.This study proposes a novel approach termed SFGAN,which embeds spatial and frequency characteristics to enhance pedestrian wind flow prediction.In the spatial domain,Gaussian blur is employed to decompose wind flow into components containing wind speed and distinguished flow edges,which are used as the embedded spatial characteristics.Detailed information of wind flow is obtained through discrete wavelet transformation and used as the embedded frequency characteristics.These spatial and frequency characteristics of wind flow are jointly utilized to enforce consistency between the predicted wind flow and ground truth during the training phase,thereby leading to enhanced predictions.Experimental results demonstrate that SFGAN clearly improves wind flow prediction,reducing Wind_MAE,Wind_RMSE and the Fréchet Inception Distance(FID)score by 5.35%,6.52%and 12.30%,compared to the previous best method,respectively.We also analyze the effectiveness of incorporating the spatial and frequency characteristics of wind flow in predicting pedestrian wind flow.SFGAN reduces errors in predicting wind flow at large error intervals and performs well in wake regions and regions surrounding buildings.The enhanced predictions provide a better understanding of performance variability,bringing insights at the early design stage to improve pedestrian wind comfort.The proposed spatial-frequency loss term is general and can be flexibly integrated with other generative models to enhance performance with only a slight computational cost.