基于SEAIQR模型与Dropout-LSTM模型的西安市COVID-19疫情趋势预测

目的基于传染病动力学SEAIQR(susceptible-exposed-asymptomatic-infected-quarantined-removed)模型和Dropout-LSTM(Dropout long short term memory network)模型预测西安市新型冠状病毒肺炎(COVID-19)疫情的发展趋势,为评估“动态清零”策略防控效果提供科学依据。方法考虑到西安市本轮疫情存在大量的无症状感染者、依时变化的参数以及采取的管控举措等特点,构建具有阶段性防控措施的时变SEAIQR模型。考虑到COVID-19疫情数据的时序性特征及它们之间的非线性关系,构建深度学习Dropout-LSTM模型。选用2021年12月9日-2022年1月31日西安市新增确诊病例数据进行拟合,用2022年2月1日-2022年2月7日数据评估预测效果,计算有效再生数(R_(t))并评价不同参数对疫情发展的影响。结果SEAIQR模型预测的新增确诊病例拐点预计在2021年12月26日出现,约为176例,疫情将于2022年1月24日实现“动态清零”,模型R^(2)=0.849。Dropout-LSTM模型能够体现数据的时序性与非线性特征,预测出的新增确诊病例数与实际情况高度吻合,R^(2)=0.937。Dropout-LSTM模型的MAE和RMSE均较SEAIQR模型低,说明预测结果更为理想。疫情暴发初期,R 0为5.63,自实施全面管控后,R_(t)呈逐渐下降趋势,直到2021年12月27日降至1.0以下。随着有效接触率不断缩小、管控措施的提早实施及免疫阈值的提高,新增确诊病例在到达拐点时的人数将会持续降低。结论建立的Dropout-LSTM模型实现了较准确的疫情预测,可为COVID-19疫情“动态清零”防控决策提供借鉴。

基于添加Dropout层的CNN-LSTM网络短期负荷预测

精准的短期负荷预测能帮助电力部门制订合理的生产调度计划,达到节省能源消耗的目的。为提升短期负荷预测的准确性,提出一种添加Dropout层的CNN-LSTM网络短期负荷预测方法。首先,根据电力负荷预测流程对预测的影响因素如气象、日期类型等进行相关性验证后构建输入特征向量;其次,使用一维卷积网络对输入的特征向量进行卷积、池化处理,并在LSTM网络全连接层添加Dropout层进行短期负荷预测仿真实验;最后,使用某电网历史数据进行测试。结果表明,相比于单独的LSTM网络,所建模型对短期负荷的预测效果更好。

基于Bi-LSTM-Dropout的蓄电池剩余使用寿命预测方法

蓄电池剩余使用寿命预测对能源的安全性和可持续发展至关重要。该文提出一种蓄电池剩余使用寿命的预测方法,利用蓄电池的历史运行数据和充放电周期,构建Bi-LSTM-Dropout网络模型。利用Bi-LSTM提取时间序列中蓄电池长期依赖的特征,采用Dropout优化算法降低Bi-LSTM网络模型的复杂度,提高模型的泛化能力。实验结果表明,该方法在测试集上的准确率达96.2%,实现了蓄电池剩余使用寿命的精确预测。

One memristor–one electrolyte-gated transistor-based high energy-efficient dropout neuronal units

Artificial neural networks(ANN) have been extensively researched due to their significant energy-saving benefits.Hardware implementations of ANN with dropout function would be able to avoid the overfitting problem. This letter reports a dropout neuronal unit(1R1T-DNU) based on one memristor–one electrolyte-gated transistor with an ultralow energy consumption of 25 p J/spike. A dropout neural network is constructed based on such a device and has been verified by MNIST dataset, demonstrating high recognition accuracies(> 90%) within a large range of dropout probabilities up to40%. The running time can be reduced by increasing dropout probability without a significant loss in accuracy. Our results indicate the great potential of introducing such 1R1T-DNUs in full-hardware neural networks to enhance energy efficiency and to solve the overfitting problem.

基于Dropout的CNN水果品质分级模型研究

【目的】提高水果分级效率。【方法】提出一种用于水果品质分级的深度学习检测识别方法,该方法在传统的CNN(卷积神经网络)模型基础上进行了改进,可以快速准确检测识别水果并进行品质分级。以苹果为例,通过对原始数据集进行数据增强,来增加样本数量和确保样本多样化。此外,为防止训练模型时出现过拟合现象,引入Dropout层来提升模型的泛化能力,同时将本研究结果与相关研究结果做对比实验分析。【结果与结论】基于Dropout的CNN水果品质分级模型对苹果品质分级精确度可达98.52%,该模型能更加准确地进行检测识别,为农业智能水果品质分级装备的研发提供了研究基础。

基于增量能量法和BiGRU-Dropout的锂电池健康状态估计

锂离子电池健康状态(SOH)的精确估计是电池管理系统面临的核心问题之一。针对实际的电池容量很难直接测量和容量再生导致的SOH估计误差问题,提出了一种基于增量能量法和双向门控循环网络(BiGRU)-Dropout的锂离子电池健康状态估计方法。首先分析增量能量曲线随电池老化的衰退规律,提取出最大峰值高度作为电池SOH的新健康因子。通过翻转层和门控循环网络层所搭建的BiGRU网络得出健康因子与SOH的映射关系,同时添加Dropout机制网络层防止出现过拟合现象,建立SOH估计模型用于电池SOH精确估计。实验结果表明,在不同充电倍率条件下,该方法均可快速、准确地估计电池SOH。

基于Dropout改进的SRGAN网络DrSRGAN

超分辨率作为一种经典的视觉任务,在多个领域有着广泛的应用。随着深度学习中无监督学习的发展,以及生成对抗网络(generative adversarial network,GAN)的提出,超分辨率技术又得到了进一步的提高,但是相关网络仍旧存在过拟合、泛化性弱等诸多问题。以超分辨率生成对抗网络(super-resolution generative adversarial network,SRGAN)为基础,受研究Dropout在经典超分辨率网络中影响的相关论文启发,在SRGAN中加入Dropout层并研究其对生成图像质量的影响,采用峰值信噪比(peak signal to noise ratio,PSNR)和结构相似性(structural similarity,SSIM)评估图像质量。实验结果表明在合适的Dropout参数下,网络重建图像具有更好的视觉效果,PSNR能够达到0.4左右的提升同时SSIM也有提高,从训练过程中不同迭代次数生成图像的比较发现改进后网络缓解了训练不稳定问题。将Dropout层加入超分辨率网络不同于以往的方法,为改进此类网络提供了一个新的思路。

基于Bi-LSTM-Dropout的新冠疫情趋势预测研究

自新冠疫情暴发以来已经对我国造成巨大影响,对新冠疫情趋势预测进行研究能够帮助人们做出应对措施。为了提高新冠疫情趋势预测模型的精度,提出了一种基于双向长短期神经网络(bi-directional long short-term memory,Bi-LSTM)建立时间序列模型。在此基础上,引入Dropout方法使神经元随机失活,解决过拟合问题。最后与长短期神经网络(long short-term memory,LSTM)、BP神经网络以及自回归积分滑动平均模型(autoregressive integrated moving average model,ARIMA)进行对比仿真实验。通过对累计确诊人数、累计死亡人数、累计治愈人数的预测结果进行对比。结果表明,文章使用的方法对于新冠疫情趋势预测明显优于其他模型。

基于集成Dropout-DNN模型的盾构掘进速度预测方法

为了提升土压平衡盾构机的掘进速度预测精度,提出一种集成Dropout技术和深度神经网络(deep neural network,DNN)模型的盾构掘进速度预测方法。依据济南地铁R1线盾构隧道段工况数据,将数据集划分为5份,并选取刀盘转速、刀盘扭矩、总推进力、螺机转速、土仓压力这5个参数为输入参数,分别建立了5个Dropout-DNN模型并进行集成实现了盾构掘进速度的预测,进一步对不同的预测方法进行了对比分析。研究结果表明:各Dropout-DNN模型预测精度具有一定的差异性,但基本良好,其决定系数均大于0.6,平均绝对百分误差均小于10%,而集成的Dropout-DNN模型决定系数为0.695,平均绝对百分误差小于5%,可见集成模型预测精度较高;基于反向传播(back propagation,BP)神经网络、DNN模型实现的盾构掘进速度预测模型其决定系数分别为0.502、0.566,可见提出的集成Dropout-DNN模型预测精度提升明显。

基于Dropout-多尺度空洞卷积神经网络的轴承故障诊断

为了提高故障诊断模型对故障轴承低信噪比信号的特征提取能力,使模型在强噪声环境下仍能发挥作用,提出了一种基于Dropout-多尺度空洞卷积神经网络(D-MDCNN)的滚动轴承故障诊断模型。首先,通过Dropout数据预处理,对训练数据进行“损坏”,强迫模型仅依靠少量特征便可进行故障诊断,以提高模型的抗噪声能力;然后,使用不同扩张率的空洞卷积扩充了多尺度信息,并利用CNN模块来完成对特征的提取与故障诊断;同时,在模型中加入批量归一化处理操作,用来加快模型训练的收敛速度,提高了模型的性能;最后,利用美国凯斯西储大学轴承数据集和东南大学齿轮箱数据集对基于D-MDCNN的模型进行了实验验证,并将实验结果与采用其他深度学习模型所得的实验结果进行了对比分析。实验结果表明:在无噪声至4dB的噪声环境下,D-MDCNN在西储大学和东南大学两个数据集上均可取得99%的诊断准确率;相比于其他同类模型,基于D-MDCNN的模型具有更高的诊断准确率和抗噪声能力。研究结果表明:基于D-MDCNN的模型是一种有效的轴承故障诊断模型。

基于Dropout优化DBN算法的数控刀具磨损状态预测

为了提高数控刀具磨损状态预测精度,建立了一种基于Dropout优化DBN方法。通过RBM训练优化网络权值与偏置量,利用误差反向算法实现网络的微调,使特征提取获得的高级特征能够良好匹配预测需求。研究结果表明:深度BP网络则获得了比BP更强学习能力,但缺乏良好的泛化能力,SVM则可以实现稳定泛化性能。当数据量持续升高后,SVM同样发生了时耗的明显增加,促进网络预测性能提升的情况下还能够更快完成特征匹配收敛过程,经过优化DBN模型表现出了更稳定与准确的预测结果。

Neural Machine Translation Models with Attention-Based Dropout Layer

In bilingual translation,attention-based Neural Machine Translation(NMT)models are used to achieve synchrony between input and output sequences and the notion of alignment.NMT model has obtained state-of-the-art performance for several language pairs.However,there has been little work exploring useful architectures for Urdu-to-English machine translation.We conducted extensive Urdu-to-English translation experiments using Long short-term memory(LSTM)/Bidirectional recurrent neural networks(Bi-RNN)/Statistical recurrent unit(SRU)/Gated recurrent unit(GRU)/Convolutional neural network(CNN)and Transformer.Experimental results show that Bi-RNN and LSTM with attention mechanism trained iteratively,with a scalable data set,make precise predictions on unseen data.The trained models yielded competitive results by achieving 62.6%and 61%accuracy and 49.67 and 47.14 BLEU scores,respectively.From a qualitative perspective,the translation of the test sets was examined manually,and it was observed that trained models tend to produce repetitive output more frequently.The attention score produced by Bi-RNN and LSTM produced clear alignment,while GRU showed incorrect translation for words,poor alignment and lack of a clear structure.Therefore,we considered refining the attention-based models by defining an additional attention-based dropout layer.Attention dropout fixes alignment errors and minimizes translation errors at the word level.After empirical demonstration and comparison with their counterparts,we found improvement in the quality of the resulting translation system and a decrease in the perplexity and over-translation score.The ability of the proposed model was evaluated using Arabic-English and Persian-English datasets as well.We empirically concluded that adding an attention-based dropout layer helps improve GRU,SRU,and Transformer translation and is considerably more efficient in translation quality and speed.

Efficient Explanation and Evaluation Methodology Based on Hybrid Feature Dropout

AI-related research is conducted in various ways,but the reliability of AI prediction results is currently insufficient,so expert decisions are indispensable for tasks that require essential decision-making.XAI(eXplainable AI)is studied to improve the reliability of AI.However,each XAI methodology shows different results in the same data set and exact model.This means that XAI results must be given meaning,and a lot of noise value emerges.This paper proposes the HFD(Hybrid Feature Dropout)-based XAI and evaluation methodology.The proposed XAI methodology can mitigate shortcomings,such as incorrect feature weights and impractical feature selection.There are few XAI evaluation methods.This paper proposed four evaluation criteria that can give practical meaning.As a result of verifying with the malware data set(Data Challenge 2019),we confirmed better results than other XAI methodologies in 4 evaluation criteria.Since the efficiency of interpretation is verified with a reasonable XAI evaluation standard,The practicality of the XAI methodology will be improved.In addition,The usefulness of the XAI methodology will be demonstrated to enhance the reliability of AI,and it helps apply AI results to essential tasks that require expert decision-making.

Guided Dropout: Improving Deep Networks Without Increased Computation

Deep convolution neural networks are going deeper and deeper.How-ever,the complexity of models is prone to overfitting in training.Dropout,one of the crucial tricks,prevents units from co-adapting too much by randomly drop-ping neurons during training.It effectively improves the performance of deep net-works but ignores the importance of the differences between neurons.To optimize this issue,this paper presents a new dropout method called guided dropout,which selects the neurons to switch off according to the differences between the convo-lution kernel and preserves the informative neurons.It uses an unsupervised clus-tering algorithm to cluster similar neurons in each hidden layer,and dropout uses a certain probability within each cluster.Thereby this would preserve the hidden layer neurons with different roles while maintaining the model’s scarcity and gen-eralization,which effectively improves the role of the hidden layer neurons in learning the features.We evaluated our approach compared with two standard dropout networks on three well-established public object detection datasets.Experimental results on multiple datasets show that the method proposed in this paper has been improved on false positives,precision-recall curve and average precision without increasing the amount of computation.It can be seen that the increased performance of guided dropout is thanks to shallow learning in the net-works.The concept of guided dropout would be beneficial to the other vision tasks.

An Ensemble Learning Model for Early Dropout Prediction of MOOC Courses

Massive open online courses(MOOCs)have become a way of online learning across the world in the past few years.However,the extremely high dropout rate has brought many challenges to the development of online learning.Most of the current methods have low accuracy and poor generalization ability when dealing with high-dimensional dropout features.They focus on the analysis of the learning score and check result of online course,but neglect the phased student behaviors.Besides,the status of student participation at a given moment is necessarily impacted by the prior status of learning.To address these issues,this paper has proposed an ensemble learning model for early dropout prediction(ELM-EDP)that integrates attention-based document representation as a vector(A-Doc2vec),feature learning of course difficulty,and weighted soft voting ensemble with heterogeneous classifiers(WSV-HC).First,A-Doc2vec is proposed to learn sequence features of student behaviors of watching lecture videos and completing course assignments.It also captures the relationship between courses and videos.Then,a feature learning method is proposed to reduce the interference caused by the differences of course difficulty on the dropout prediction.Finally,WSV-HC is proposed to highlight the benefits of integration strategies of boosting and bagging.Experiments on the MOOCCube2020 dataset show that the high accuracy of our ELM-EDP has better results on Accuracy,Precision,Recall,and F1.

DropoutVNS:基于变领域搜索的dropout方法

为了揭示和改进标准的dropout,使用变领域搜索算法(Variable Neighborhood Search,VNS)的思想提出一种新的解释,将dropout看作是一种特殊的变领域搜索算法,训练不同的网络相当于在变化领域搜索最优解。同时提出一种新的正则化方法:dropoutVNS (Dropout Method Based On Variable Neighborhood Search)。DropoutVNS的核心思想是改变dropout切换网络的策略,将完全随机切换转变为稳定时再切换,以改善标准的dropout。在三个常用的图像实验集的实验结果表明,dropoutVNS可以减少训练时间,有效提高模型在图像分类上的预测准确率。

基于Dropout正则化的汉语框架语义角色识别

汉语框架语义角色识别是汉语框架语义分析的重要任务之一。该文基于汉语词语、词性等特征的分布式表示,使用一种多特征融合的神经网络结构来构建汉语框架语义角色识别模型。鉴于可用的训练语料规模有限,该文采用了Dropout正则化技术来改进神经网络的训练过程。实验结果表明,Dropout正则化的加入有效地缓解了模型的过拟合现象,使得模型的F值有了近7%的提高。该文进一步优化了学习率以及分布式表示的初始值,最终的汉语框架语义角色识别的F值达到70.54%,较原有的最优结果提升2%左右。

连续语音识别中基于Dropout修正线性深度置信网络的声学模型

大词汇量连续语音识别系统中,为了增强现有声学模型的表征能力、防止模型过拟合,提出一种基于遗失策略(Dropout)修正线性深度置信网络的声学模型构建方法。该方法使用修正线性函数代替传统Logistic函数进行深度置信网络训练,修正线性函数更接近生物神经网络的工作方式,增强了模型的表征能力;同时引入Dropout策略对修正线性深度置信网络进行调整,避免节点之间的协同作用,防止网络出现过拟合。文章利用公开语音数据集进行了实验,实验结果证明了所提出的声学模型构建方法相对于传统方法的优越性。

运用Dropout-LSTM模型的新冠肺炎趋势预测

为提高新冠肺炎(COVID-19)趋势预测精度,该文提出一种运用Dropout技术的长短期记忆(LSTM)神经网络预测新冠肺炎发展趋势的方法。该方法基于Python语言使用网络爬虫技术获取完整的国内新冠肺炎历史数据,提高数据采集效率的同时减少了主观原因导致的数据错误;因为新冠肺炎历史数据为时序性数据,为避免人为添加时间特征及充分挖掘较少时序数据之间的非线性关系,该文构建了层数更多的LSTM神经网络预测模型。随后在隐藏层中的非循环部分采用Dropout技术,对神经元进行随机概率失活,有效解决了深度学习的过拟合问题。最后用国内累计确诊、现有确诊和累计治愈人数对该方法进行验证,实验证明该方法可较精准预测新冠肺炎传播趋势。

基于ANN-dropout的配电网可靠性预测方法

随着配电网的高速发展,大量自动化设备不断接入使用给配电网的可靠运行带来了巨大的挑战。本文针对配电网动态规律和可靠性指标特征,提出了两种机器学习算法来预测配电网可靠性。通过对配电网数据进行数据变换,让预测区间从[0,1]变成[0,+∞),对变换后的数据进行归一化再输入预测模型。配电网实例验证结果表明,数据变换后机器的学习预测效果提升明显,经过dropout技术优化的人工神经网络(artificial neural network,ANN)具有最佳的预测性能。本文提出的预测模型能准确地预测配电网的可靠性,为配电网的建设投资和优化运行提供有效的指导。