Echo State Network With Probabilistic Regularization for Time Series Prediction

Recent decades have witnessed a trend that the echo state network(ESN)is widely utilized in field of time series prediction due to its powerful computational abilities.However,most of the existing research on ESN is conducted under the assumption that data is free of noise or polluted by the Gaussian noise,which lacks robustness or even fails to solve real-world tasks.This work handles this issue by proposing a probabilistic regularized ESN(PRESN)with robustness guaranteed.Specifically,we design a novel objective function for minimizing both the mean and variance of modeling error,and then a scheme is derived for getting output weights of the PRESN.Furthermore,generalization performance,robustness,and unbiased estimation abilities of the PRESN are revealed by theoretical analyses.Finally,experiments on a benchmark dataset and two real-world datasets are conducted to verify the performance of the proposed PRESN.The source code is publicly available at https://github.com/LongJinlab/probabilistic-regularized-echo-state-network.

Echo state network based symbol detection in chaotic baseband wireless communication

The Chaotic Baseband Wireless Communication System(CBWCS)is expected to eliminate the Inter-Symbol Interference(ISI)caused by multipath propagation by using the optimal decoding threshold that is the sum of the ISI caused by past decoded bits and the ISI caused by future transmitting bits.However,the current technique is only capable of removing partial effects of the ISI,because only past decoded bits are available for the suboptimal decoding threshold calculation.The unavailability of the future information needed for the optimal decoding threshold is an obstacle to further improve the Bit Error Rate(BER)performance.In contrast to the previous method using Echo State Network(ESN)to predict one future bit,the proposed method in this paper predicts the optimal decoding threshold directly using ESN.The proposed ESN-based threshold prediction method simplifies the symbol decoding operation by avoiding the iterative prediction of the output waveform points using ESN and accumulated error caused by the iterative operation.With this approach,the calculation complexity is reduced compared to the previous ESN-based approach.The proposed method achieves better BER performance compared to the previous method.The reason for this superior result is twofold.First,the proposed ESN is capable of using more future symbols information conveyed by the ESN input to obtain more accurate threshold rather than the previous method in which only one future symbol was available.Second,the proposed method here does not need to estimate the channel information using Least Squared(LS)method,which avoids the extra error caused by inaccurate channel information estimation.Simulation results and experiment based on a wireless open-access research platform under a practical wireless channel show the effectiveness and superiority of the proposed method.

基于Echo State Neural Networks的短期交通流预测算法

在城市交通环境,交通流的正确预测是比较困难,因为多个十字路口,这使得预置的交通控制模型之间的相互作用和intertwinement不能保持始终高性能在所有的交通情况。

Simplified Echo-State-Network Based Services Awareness for High-Speed Passive Optical Network

With the challenge from services diversity grows greatly,the service-oriented supporting ability is required to current high-speed passive optical network(PON) .Aimed to enhance the quality of service(Qo S) brought by diversified-services,this paper proposes an Simplified Echo State Network(SESN) Based Services Awareness scheme in High-Speed PON(Passive Optical Network) .In this proposed scheme,the ring topology is adopted in the reservoir of SESN to reduce the complexity of original Echo State Network,and system dynamics equation is introduced to keep the accuracy of SESN.According to the network architecture of 10G-EPON,a SESN Master is running in the OLT and a number of SESN Agents work in ONUs.The SESN Master plays the main function of service-awareness from the total view of various kinds services in 10G-EPON system,by fully SESN training.Then,the reservoir information of well-trained SESN in OLT will be broadcasted to all ONUs and those SESN Agents working in ONUs are allowed to conducts independent service-awareness function.Thus,resources allocation and transport policy are both determined just only in ONUs.Simulation results show that the proposed mechanism is able to better supporting ability for multiple services.

Stock Price Forecasting: An Echo State Network Approach

Forecasting stock prices using deep learning models suffers from pro-blems such as low accuracy,slow convergence,and complex network structures.This study developed an echo state network(ESN)model to mitigate such pro-blems.We compared our ESN with a long short-term memory(LSTM)network by forecasting the stock data of Kweichow Moutai,a leading enterprise in China’s liquor industry.By analyzing data for 120,240,and 300 days,we generated fore-cast data for the next 40,80,and 100 days,respectively,using both ESN and LSTM.In terms of accuracy,ESN had the unique advantage of capturing non-linear data.Mean absolute error(MAE)was used to present the accuracy results.The MAEs of the data forecast by ESN were 0.024,0.024,and 0.025,which were,respectively,0.065,0.007,and 0.009 less than those of LSTM.In terms of con-vergence,ESN has a reservoir state-space structure,which makes it perform faster than other models.Root-mean-square error(RMSE)was used to present the con-vergence time.In our experiment,the RMSEs of ESN were 0.22,0.27,and 0.26,which were,respectively,0.08,0.01,and 0.12 less than those of LSTM.In terms of network structure,ESN consists only of input,reservoir,and output spaces,making it a much simpler model than the others.The proposed ESN was found to be an effective model that,compared to others,converges faster,forecasts more accurately,and builds time-series analyses more easily.

A Prediction Method Based on Improved Echo State Network for COVID-19 Nonlinear Time Series

This paper proposes a prediction method based on improved Echo State Network for COVID-19 nonlinear time series, which improves the Echo State Network from the reservoir topology and the output weight matrix, and adopt the ABC (Artificial Bee Colony) algorithm based on crossover and crowding strategy to optimize the parameters. Finally, the proposed method is simulated and the results show that it has stronger prediction ability for COVID-19 nonlinear time series.

Echo-state-network classification based multi-services awareness in high-speed optical passive networks

With the challenge of great growing of services diversity,service-oriented supporting ability is required by current high-speed passive optical network( PON). Aimed at enhancing the quality of service( Qo S) brought by diversified-services,this study proposes an echo state network( ESN)based multi-service awareness mechanism in 10-Gigabite ethernet passive optical network( 10GEPON). In the proposed approach,distributed architecture is adopted to realize this ESN based multi-service awareness. According to the network architecture of 10G-EPON,where a main ESN is running in OLT and a number of ESN agents works in ONUs. The main-ESN plays the main function of service-awareness from the total view of various kinds of services in 10G-EPON system,by full ESN training. Then,the reservoir information of well-trained ESN in OLT will be broadcasted to all ONUs and those ESN agents working in ONUs are allowed to conduct independent service-awareness function. Thus,resources allocation and transport policy are both determined only in ONUs. Simulation results show that the proposed mechanism is able to better support the ability of multiple services.

基于变分模态分解与鲸鱼算法优化回声状态网络的风速预测模型

风速受多种因素影响常伴随着随机性和非平稳性,给风电接入电网造成了相当大的困难,准确的风速预测对风力发电有着极大的研究意义。将变分模态分解算法与鲸鱼算法优化回声状态网络模型相结合,提出了一种风速预测模型。首先通过变分模态分解算法将风速序列分解成多个分量以减少风速内部信号间的耦合性,降低建模难度。然后对这些分量分别建立对应的回声状态网络预测模型。针对回声状态网络模型性能受储备池参数影响较大的问题,采用鲸鱼优化算法对储备池参数进行优化。风速的最终预测值由分解后各分量预测值相加得到。最后,将实际采集的短期风速数据作为研究对象,通过与其他4种预测模型的对比分析表明提出的风速预测模型具有更高的预测精度,能够更好地对风速的变化趋势进行预测。

基于TCN-BiLSTM-Attention-ESN的光伏功率预测

针对光伏发电功率随机性强、难以准确预测的问题,提出一种基于时间卷积网络(TCN)、双向长短期记忆网络(BiLSTM)和回声状态网络(ESN)的组合预测方法。首先,使用自适应噪声完备集合经验模态分解(CEEMDAN)将功率数据分解为一系列相对平稳、不同波动模式的子功率序列;再将分解重构后的功率序列和其他特征序列输入到TCN-BiLSTM-Attention-ESN组合模型中,其中TCN-BiLSTM-Attention用于提取光伏序列波动特征并构建时空特征向量;最后,将所提取的时空特征向量输入ESN获得预测结果。采用新疆某光伏电站的光伏功率数据进行验证,结果表明与时下先进的预测方法相比,所提方法具有更高的预测精度,有助于提升光伏发电占比,保障电力系统平衡和运行安全。

偏置剪枝叠式自编码回声状态网络的时序预测

针对大多数模型对时间序列预测数据的预测准确率较低,为提升时间序列的预测精度,提出一种基于Biased Drop-weight的偏置剪枝叠式自编码回声状态网络(BD-AE-SGESN)的深度模型。以叠式ESN为多层深度网络框架,提出一种生成式AE算法生成每一层的输入权值,利用BD算法根据输入权重激活值进行剪枝。对比实验结果表明,该模型能够有效提升预测准确率,在3个不同的数据上,相比其它模型有着较小的预测误差和较高的稳定度。

基于蝙蝠算法优化ESN的氯乙烯质量分数软测量模型预测

为解决氯乙烯因其精馏过程具有较强的非线性,无法实现对氯乙烯质量分数实时测量的问题,提出一种基于蝙蝠算法(bat algorithm,BA)优化回声状态网络(echo state network,ESN)的软测量模型BA-ESN.首先,通过对氯乙烯精馏过程的分析,选取模型的辅助变量,并将归一化处理后的数据作为模型输入变量;其次,由于回声状态网络中的权值和阈值都是随机产生的,影响其泛化能力,故采用蝙蝠算法对回声状态网络的输出权值进行优化,从而提高ESN模型的收敛速度;最后,将BA-ESN模型预测氯乙烯质量分数的预测结果与ESN模型和BP模型的预测结果进行对比.仿真结果表明:BA-ESN模型的预测精度较高,泛化能力和鲁棒性都较好,能够满足氯乙烯精馏过程实时测量的要求.

基于知识与AW-ESN融合的烧结过程FeO含量预测

氧化亚铁(FeO)含量是衡量烧结矿强度和还原性的重要指标,烧结过程FeO含量的实时准确预测对于提升烧结质量、优化烧结工艺具有重要意义.然而烧结过程热状态参数缺失、过程参数波动频繁给FeO含量的高精度预测带来巨大的挑战,为此,提出一种基于知识与变权重回声状态网络融合(Fusion of data-knowledge and adaptive weight echo state network, DK-AWESN)的烧结过程FeO含量预测方法.首先,针对烧结过程热状态参数缺失的问题,建立烧结料层最高温度分布模型,实现基于料层温度分布特征的FeO含量等级划分;其次,针对烧结过程参数波动频繁的问题,提出基于核函数高维映射的多尺度数据配准方法,有效抑制离群点的影响,提升建模数据的质量;最后,针对烧结过程数据驱动模型缺乏机理认知致使模型预测精度不高的问题,将过程数据中提取得到的FeO含量等级知识与AW-ESN (Adaptive weight echo state network)结合,建立DK-AWESN模型,有效提升复杂工况下FeO含量的预测精度.现场工业数据试验表明,所提方法能实时准确地预测烧结过程FeO含量,为烧结过程的智能化调控提供实时有效的FeO含量反馈信息.

基于PLESN和LESQRN概率预测模型的短期电力负荷预测

针对现有电力负荷预测不能很好反映负荷数据的周期性和趋势性以及残差的波动性特征提出一种考虑周期性建模的泄露积分型回声状态网络点预测模型和泄露积分型回声状态分位数回归网络概率预测模型组合的短期电力负荷预测方法.首先为了捕捉负荷的多重特征定义了周期性和趋势性损失函数辅助优化点预测模型然后为克服残差的波动问题利用概率预测模型对点预测值与真实值的残差进行建模预测最后整合同时刻的点预测值与残差预测区间得到概率预测模型结果.实际算例结果表明与其他模型相比所提模型不仅有效抑制尖端振荡现象而且能够生成可靠的概率密度分布.

储能用质子交换膜燃料电池长期老化预测

质子交换膜燃料电池(PEMFC)的长期老化预测有助于缩短耐久性测试时间,降低成本,为维护策略提供依据。针对超参数问题,提出一种将优化算法和储备池计算相结合的数据驱动预测方法。基于耐久性测试数据集,以电堆输出电压为老化指标,利用麻雀搜索算法(SSA)优化回声状态网络(ESN)的储备池尺寸、泄漏率和正则化系数,以构建预测模型。分别利用原始数据的前30%、40%、50%和60%作为训练集训练模型,验证模型在各训练集比例下的长期老化预测性能。训练集比例为30%时,所提方法在静态工况下的长期预测均方根误差(RMSE)达到0.0083,准动态工况下可达到0.0359。

基于EMD-DESN的无人机集群航迹目的地预测

无人机(unmanned aerial vehicle,UAV)集群作战样式多样、运动模式复杂,导致集群航迹目的地难以预测。为解决上述问题,本文提出了一种基于经验模态分解(empirical mode decomposition,EMD)和深度回声状态网络(deep echo state network,DESN)的UAV集群航迹目的地预测算法。为使集群运动模型更真实地模拟UAV集群作战过程,本文引入航向误差时变方差,改进了Olfati-Saber集群运动模型的虚拟领导项。为处理因群内的协同作用和集群航向误差导致的运动非平稳性,引入了EMD,对UAV航迹序列进行重构。考虑到获知航迹的时序性,设计了滑窗结构,采用DESN对重构航迹的不同时段进行目的地预测。仿真实验结果表明,本文提出的EMD-DESN算法较基本DESN算法能以更高的准确度预测UAV集群航迹目的地,并能更早地实现稳定的正确预测。

基于双通道回声状态网络的时间序列补全及单步预测

随着物联网的发展,众多传感器采集到大量具有丰富数据相关性的时间序列,为各种数据挖掘应用提供强大的数据支持。然而,一些客观或主观原因(如设备故障、稀疏感知等)往往会造成采集到的数据出现不同程度的缺失。虽然已有很多方法被提出用于解决这一问题,但这些方法在数据相关性方面或考虑不够全面,或计算成本过高。而且,现有方法仅关注对缺失值的补全,未能兼顾下游应用。针对上述不足,设计了一种兼顾补全与预测任务的双通道回声状态网络。两个通道的网络虽共用输入层,但具有各自的储备池和输出层。两者最大的区别是左/右通道的输出层分别表示输入层前/后一个时刻对应的目标值或预补值。最后将两个通道的估计值进行融合,充分利用来自缺失时刻之前和之后的数据相关性以进一步提升性能。两种缺失现象下(随机缺失和分段缺失)不同缺失率的实验结果表明,所提模型无论是在补全精度还是预测精度上都优于目前流行的各类方法。

基于战争策略算法优化回声状态网络的时间序列预测

为了解决回声状态网络(ESN)储备池参数难以确定的问题,提出一种基于战争策略优化算法(WSO)的回声状态网络模型(WSO-ESN).该模型利用战争策略优化算法中攻击和防御两种流行的战争策略更好地实现整个模型在全局探索和局部开发上的平衡,替换弱士兵策略提高其鲁棒性使WSO算法在确定ESN参数时更准确.此外,还引入了呈指数变化的权重更新机制提高算法的收敛速度进而更快地确定储备池参数.实验结果与粒子群优化算法(PSO)、蜣螂优化算法(DBO)、金豺优化算法(GJO)等对储备池参数优化方法进行比较.结果表明,基于战争策略优化算法的回声状态网络模型具有更快的训练速度和更高的预测精度.

改进DBO优化CRJ网络的PEMFC剩余使用寿命预测

质子交换膜燃料电池(PEMFC)剩余使用寿命(RUL)预测是其健康管理和故障诊断的主要问题。针对此问题提出一种基于改进蜣螂算法(HDBO)优化确定性循环跳跃储备池网络(CRJ)的PEMFC剩余使用寿命预测方法。采用局部加权回归散点平滑法对PEMFC运行数据进行重构和平滑处理,有效地滤除噪声并保存了原始数据的主要特征。利用最大信息系数(MIC)结合贝叶斯信息准则(BIC)方法选择出8个最优输入特征。采用混沌映射初始化种群、自适应调整搜索策略、引入柯西变异改进了蜣螂优化算法,利用HDBO优化CRJ的三个关键参数,建立高效的预测模型。将最优特征集作为预测模型的输入实现PEMFC的剩余使用寿命预测,实验结果表明,该方法的决定系数(R2)、平均绝对值误差(MAE)和均方根误差(RMSE)分别为0.95022、0.0025729和0.0035232,与DBO、SSA和CRJ相比,该方法的预测精度更高。

基于图正则化自编码回声状态网络的时间序列分类算法

回声状态网络(Echo State Network,ESN)能够为解决时间序列问题提供有效的动态解决方法,然而大多数情况下ESN模型主要用于预测而不是分类,ESN在时间序列分类任务的应用尚未得到充分的研究。传统ESN由于存在随机生成的输入权重,使得其性能并不能保证最优。随机生成的权重在特征映射时,可能会破坏有用的特征。针对这些缺点,提出了一种针对时间序列分类任务的基于图正则化自编码的回声状态网络模型(GRAE-ESN),利用流形学习考虑数据内在的流形结构,来约束输出权重使得相似样本的输出在新的空间中更加接近,之后将ESN结构中的输入权重用解码层获得的权重所替换,以学习到丰富的输入特征。在基准数据上的实验表明,所提出的GRAE方法能够有效的改进ESN分类器,在与多个主流算法和深度学习算法相比,该算法具有更好的性能和鲁棒性。

Lightweight Intrusion Detection Using Reservoir Computing

The blockchain-empowered Internet of Vehicles(IoV)enables various services and achieves data security and privacy,significantly advancing modern vehicle systems.However,the increased frequency of data transmission and complex network connections among nodes also make them more susceptible to adversarial attacks.As a result,an efficient intrusion detection system(IDS)becomes crucial for securing the IoV environment.Existing IDSs based on convolutional neural networks(CNN)often suffer from high training time and storage requirements.In this paper,we propose a lightweight IDS solution to protect IoV against both intra-vehicle and external threats.Our approach achieves superior performance,as demonstrated by key metrics such as accuracy and precision.Specifically,our method achieves accuracy rates ranging from 99.08% to 100% on the Car-Hacking dataset,with a remarkably short training time.