基于长短时记忆网络的人体姿态检测方法

针对在循环神经网络（RNN）网络结构下较为遥远的历史信号无法传递至当前时刻的问题,长短时记忆（LSTM）网络作为RNN的一种变体被提出,在继承RNN对时间序列优秀的记忆能力的前提下,LSTM克服了这种时间序列的长期依赖问题,并在自然语言处理与语音识别领域有较好的表现。对于人体行为动作中也存在作为时间序列的长期依赖问题与使用传统滑窗算法采集数据时造成的无法实时检测的问题,将LSTM扩展应用到人体姿态检测,提出了基于LSTM的人体姿态检测方法。通过目前智能手机中一般都带有的加速度传感器、陀螺仪、气压计和方向传感器实时采集的时序数据,制作了包含3336条带有人工标注数据的人体姿态数据集,对行走、奔跑、上楼梯、下楼梯和平静五种日常持续性行为姿态与跌倒、起立、坐下和跳跃这四个突发行为姿态进行预测分类。对比LSTM网络与该研究领域内常用的浅层学习算法、深度学习全连接神经网络与卷积神经网络,实验结果表明,所提方法使用端对端的深度学习的方法相比基于所制作数据集的人体姿态检测算法模型的正确率提高了4.49个百分点,验证了该网络结构的泛化能力且更适合姿态检测。     

非饱和区扩展的RNN算法优化

针对长短时记忆网络(LSTM)型循环神经网络(RNN)收敛速度慢,提出了扩展激活函数非饱和区的RNN算法优化.针对LSTM型RNN的训练过程收敛速度慢的原因以及激活函数的性质,提出了加快RNN训练过程收敛的解决方法.通过字符级语言模型对优化方法进行了验证,结果表明:非饱和区扩展的RNN算法优化有效地加快了RNN训练过程的收敛.     

基于遗传算法改进的一阶滞后滤波和长短期记忆网络的蓝藻水华预测方法

河湖藻类水华形成过程中所具有的突发性和不确定性,导致对藻类水华爆发预测准确性不高。为解决此问题,以叶绿素a的浓度值作为蓝藻水华演化过程表征指标,提出基于长短期记忆（LSTM）循环神经网络（RNN）蓝藻水华预测模型。首先,用遗传算法改进的一阶滞后滤波（GF）优化算法对数据进行平滑滤波处理;然后,搭建GF-LSTM网络的蓝藻水华预测模型,实现对水华发生的精准预测;最后,以太湖水域梅梁湖区域的采样数据为样本,对预测模型进行检验,并与传统的RNN和LSTM网络进行对比。仿真结果表明,提出的GF-LSTM网络模型平均相对误差控制在16%~18%,而RNN模型的预测平均相对误差为28%~32%,LSTM网络模型的平均相对误差为19%~22%,对采用数据的平滑性处理效果较好,预测精度更高,对样本具有更好的适应性,克服了传统RNN模型在长期训练时出现的梯度消失与梯度爆炸缺点。     

最小窥视孔长短时记忆模型

由于循环神经网络拥有复杂的模型结构,使训练模型达到最优变得困难。因此,提出一种最小窥视孔长短时记忆模型,它只有一个唯一门来更新信息,拥有两个网络层,通过减少一定的模型参数降低模型训练的难度,提高模型性能。实验结果表明,在不同数据集上,该模型性能高于长短期记忆模型,部分高于门循环单元模型,在参数个数、运行时间方面,其远小于长短期记忆模型以及门循环单元模型。     

基于Bi-LSTM的财金文本情感分类研究

针对现有文本情感分析方法准确率不高、实时性不强以及特征提取不充分的问题,构建了双向长短时记忆神经网络(Bi-LSTM)模型。利用双向循环长短时记忆神经网络(Bi-LSTM)提取文本的上下文信息;然后,对已提取的上下文特征利用LSTM进行局部语义特征提取;最后,采用语法规则,构建情感分类器。通过与RNN、Bi-LSTM等单一模型对比。实验结果表明,多层LSTM结构相比传统RNN(循环神经网络)的ROUGE(语义恢复度)值更高,具有较好的学习能力。     

Optimal PID control for hovering stabilization of quadcopter using long short term memory

Drones are a type of unmanned aerial vehicle. They use several rotors to control their flight motion and stabilize their attitude. This study aims to determine the optimal proportional-integral–differential (PID) gain values that can stabilize a quadcopter with four rotors quickly when its attitude is disturbed. Generally, expert knowledge and a great deal of time and money are required to obtain the PID gain values to stabilize the attitude of a drone. However, in this study, long short-term memory (LSTM), which is a type of neural network algorithm, was used to evaluate the flight motions based on PID gain values without expert knowledge of quadcopters (quadcopter flight motion, PID control algorithms, and quadcopter expert experience). To obtain the optimal values of the PID gain for stabilizing the attitude of a drone, a PID simulator algorithm was developed in this study. The developed algorithm used dynamic equations of motion for the drones. Simulations were used to acquire the drone stabilization data, and a back-propagation neural network was applied to establish an approximate model. Subsequently, the non-dominated sorting genetic algorithm-II was used to obtain the optimal PID gain values that could restore the attitude of the drone quickly. The drone motion data obtained using the simulations were used as the LSTM training data, and the optimal PID gain values obtained using the genetic algorithm were used by the LSTM to predict the motion. To verify the results, a drone was constructed, and the LSTM and dynamic simulation values were compared with the drone experimental values using an experimental device that allowed the motion of the drone to be examined. The results indicated that the motions resulting from the optimal values and the experimental results were in agreement.     

K-Similarity降噪的LSTM神经网络水质多因子预测模型

针对水质预测问题,以地表水水质监测因子作为研究对象,提出了一种基于长短期记忆（LSTM）神经网络的水质多因子预测模型,同时利用提出的K-Similarity降噪法对模型的输入数据进行降噪,提高模型预测性能.通过与BP神经网络、RNN和传统的LSTM神经网络预测模型进行对比实验,证明了所提出的方法均方误差最小,预测结果更准确.     

Using long short-term memory recurrent neural network in land cover classification on Landsat and Cropland data layer time series

Land cover maps are significant in assisting agricultural decision making. However, the existing workflow of producing land cover maps is very complicated and the result accuracy is ambiguous. This work builds a long short-term memory (LSTM) recurrent neural network (RNN) model to take advantage of the temporal pattern of crops across image time series to improve the accuracy and reduce the complexity. An end-to-end framework is proposed to train and test the model. Landsat scenes are used as Earth observations, and some field-measured data together with CDL (Cropland Data Layer) datasets are used as reference data. The network is thoroughly trained using state-of-the-art techniques of deep learning. Finally, we tested the network on multiple Landsat scenes to produce five-class and all-class land cover maps. The maps are visualized and compared with ground truth, CDL, and the results of SegNet CNN (convolutional neural network). The results show a satisfactory overall accuracy (> 97% for five-class and > 88% for all-class) and validate the feasibility of the proposed method. This study paves a promising way for using LSTM RNN in the classification of remote sensing image time series.     

Times-series data augmentation and deep learning for construction equipment activity recognition

Automated, real-time, and reliable equipment activity recognition on construction sites can help to minimize idle time, improve operational efficiency, and reduce emissions. Previous efforts in activity recognition of construction equipment have explored different classification algorithms anm accelerometers and gyroscopes. These studies utilized pattern recognition approaches such as statistical models (e.g., hidden-Markov models); shallow neural networks (e.g., Artificial Neural Networks); and distance algorithms (e.g., K-nearest neighbor) to classify the time-series data collected from sensors mounted on the equipment. Such methods necessitate the segmentation of continuous operational data with fixed or dynamic windows to extract statistical features. This heuristic and manual feature extraction process is limited by human knowledge and can only extract human-specified shallow features. However, recent developments in deep neural networks, specifically recurrent neural network (RNN), presents new opportunities to classify sequential time-series data with recurrent lateral connections. RNN can automatically learn high-level representative features through the network instead of being manually designed, making it more suitable for complex activity recognition. However, the application of RNN requires a large training dataset which poses a practical challenge to obtain from real construction sites. Thus, this study presents a data-augmentation framework for generating synthetic time-series training data for an RNN-based deep learning network to accurately and reliably recognize equipment activities. The proposed methodology is validated by generating synthetic data from sample datasets, that were collected from two earthmoving operations in the real world. The synthetic data along with the collected data were used to train a long short-term memory (LSTM)-based RNN. The trained model was evaluated by comparing its performance with traditionally used classification algorithms for construction equipment activity recognition. The deep learning framework presented in this study outperformed the traditionally used machine learning classification algorithms for activity recognition regarding model accuracy and generalization.     

Combining empirical mode decomposition and deep recurrent neural networks for predictive maintenance of lithium-ion battery

Predictive maintenance of lithium-ion batteries has been one of the popular research subjects in recent years. Lithium-ion batteries can be used as the energy supply for industrial equipment, such as automated guided vehicles and battery electric vehicles. Predictive maintenance plays an important role in the application of smart manufacturing. This mechanism can provide different levels of pre-diagnosis for machines or components. Remaining useful life (RUL) prediction is crucial for the implementation of predictive maintenance strategies. RUL refers to the estimated useful life remaining before the machine cannot operate after a certain period of operation. This study develops a hybrid data science model based on empirical mode decomposition (EMD), grey relational analysis (GRA), and deep recurrent neural networks (RNN) for the RUL prediction of lithium-ion batteries. The EMD and GRA methods are first adopted to extract the characteristics of time series data. Then, various deep RNNs, including vanilla RNN, gated recurrent unit, long short-term memory network (LSTM), and bidirectional LSTM, are established to forecast state of health (SOH) and the RUL of lithium-ion batteries. Bayesian optimization is also used to find the best hyperparameters of deep RNNs. Experimental results with the lithium-ion batteries data of NASA Ames Prognostics Data Repository show that the proposed hybrid data science model can accurately predict the SOH and RUL of lithium-ion batteries. The LSTM network has the optimal results. The proposed hybrid data science model with multiple artificial intelligence-based technologies also demonstrates critical digital-technology enablers for digital transformation of smart manufacturing and transportation.     

Minimal gated unit for recurrent neural networks

Recurrent neural networks (RNN) have been very successful in handling sequence data. However, understanding RNN and finding the best practices for RNN learning is a difficult task, partly because there are many competing and complex hidden units, such as the long short-term memory (LSTM) and the gated recurrent unit (GRU). We propose a gated unit for RNN, named as minimal gated unit (MGU), since it only contains one gate, which is a minimal design among all gated hidden units. The design of MGU benefits from evaluation results on LSTM and GRU in the literature. Experiments on various sequence data show that MGU has comparable accuracy with GRU, but has a simpler structure, fewer parameters, and faster training. Hence, MGU is suitable in RNN's applications. Its simple architecture also means that it is easier to evaluate and tune, and in principle it is easier to study MGU's properties theoretically and empirically.     

基于LSTM的疏浚管道流速预测模型

疏浚管道输送系统是绞吸挖泥船的重要组成部分,在施工过程中对流速的控制至关重要。该文以现有疏浚管道输送实验台为对象,提出了一种疏浚管道输送系统的流速预测模型。首先针对反向传播神经网络(BP)无法处理序列间的关联信息以及传统循环神经网络(RNN)无法记忆久远关键信息的缺陷,提出了基于长短期记忆循环神经网络(LSTM)的流速预测模型;然后使用LSTM模型对疏浚管道输送实验台和绞吸挖泥船的数据集进行网络训练,并对管道流速进行预测。通过将流速的预测值与真实值进行对比,验证了该文提出的LSTM模型具有很强的适用性和很高的准确性。     

基于CNN-LSTM的活动识别方法

在识别活动时,传统的循环神经网络RNN识别方法不考虑传感器活动数据之间依赖性强的问题,导致识别准确率降低。为了提高识别准确率,解决活动数据依赖性强的问题,用长短期记忆网络LSTM进行活动识别,LSTM在考虑当前点输入的同时考虑先前点的输出,能够保持数据之间的强依赖性。但是,LSTM在处理传感器活动数据的特征提取方面时间效率不高,而卷积神经网络CNN能共享卷积核,且可以从杂乱无章的数据中提取出明显特征向量。提出一种基于CNN-LSTM的活动识别方法CLAR,利用CNN能够很好地提取出活动序列数据中的特征向量,并将提取出的特征向量作为LSTM的输入,利用LSTM门限之间的相互作用进行活动识别,使得依赖性很强的活动数据成为活动识别的优势,进而提高活动识别的准确率和时间效率。实验表明,CLAR方法的识别准确率比单一神经网络活动识别方法的准确率提高了9%,时间平均缩短了10%。     

SDAE-LSTM模型在金融时间序列预测中的应用

针对金融时间序列预测的复杂性和长期依赖性，提出了一种基于深度学习的LSTM神经网络预测模型。利用堆叠去噪自编码从金融时间序列的基本行情数据和技术指标中提取特征，将其作为LSTM神经网络的输入对金融时间序列进行预测；通过LSTM神经网络的长期依赖特性来提高金融时间序列的预测精度。利用股价指数数据，与传统的神经网络的预测结果进行比较，结果表明基于深度学习的LSTM神经网络具有比较高的预测精度。     

基于多流CNN-LSTM网络的群体情绪识别

群体情绪识别是人机交互领域的前言课题,针对群体情绪识别准确率的问题,结合卷积神经网络（CNN）与长短期记忆网络（LSTM）,提出一种多流CNN-LSTM网络模型学习群体情绪的静态和动态特征。以视频序列的原始图像、视觉显著图形和叠加的光流图像分别作为三个通道的输入,利用CNN网络对空间特征和局部运动特征进行分析,得到的特征图直接输入LSTM网络,进行全局运动特征的学习。最后连接Softmax分类器,对三个通道的Softmax输出进行加权融合,得到分类结果。实验结果表明,本文模型可有效地识别4种典型的群体情绪,且识别率高于已有算法,准确度（ACC）和宏平均精度（MAP）分别最高可达82.6%、84.1%。     

基于多元时间序列分析的控制系统执行器故障诊断方法研究

为提高控制系统执行器故障实时诊断的准确率,该文提出一种基于多元时间序列分析的控制系统执行器在线故障诊断方法。首先分析了控制系统执行器故障机理,确定了表征执行器故障的关键信号;其次采用执行器历史数据,建立了时间卷积网络(TCN)在线预测模型,对执行器多通道信号进行在线预测;随后通过长短期记忆网络(LSTM)对多通道残差信号建立了故障分类模型;最后以燃气轮机控制系统执行器半物理试验平台中的电液执行器为例进行了多次重复试验验证。结果表明,基于TCN网络的在线预测模型相比传统循环神经网络(RNN)预测误差较小;基于LSTM网络的故障分类模型准确率较高;通过LSTM网络对多通道残差信号进行故障分类,比对原始故障数据分类故障准确率更高。     

基于深度学习的人体动作识别方法

对基于机器视觉的人体动作识别的成果进行研究,为提高视频数据集中人体动作的识别率,提出一种改进的深度网络模型。采用稠密光流方法处理数据,结合二维卷积神经网络(2DCNN)、三维卷积神经网络(3DCNN)和长短期记忆神经网络(LSTM)对动作特征进行提取,利用Softmax分类器识别分类。通过KTH数据集进行实验对比验证,分析结果表明,改进模型相比其它已有模型具有更高的识别率,动作识别效果更优。     

基于极性转移和LSTM的树结构网络与句子分类

长短期记忆网络(long short term memory,LSTM)是一种能长久储存序列信息的循环神经网络,在语言模型、语音识别、机器翻译等领域都得到了广泛的应用。先研究了前人如何将LSTM中的记忆模块拓展到语法树得到LSTM树结构网络模型,以获取和储存句子深层次的语义结构信息;然后针对句子词语间的极性转移在LSTM树结构网络模型中添加了极性转移信息提出了极性转移LSTM树结构网络模型,更好获取情感信息来进行句子分类。实验表明在Stanford sentiment tree-bank数据集上,提出的极性转移LSTM树结构网络模型的句子分类效果优于LSTM、递归神经网络等模型。     

Learning to forget: continual prediction with LSTM

Long short-term memory (LSTM; Hochreiter & Schmidhuber, 1997) can solve numerous tasks not solvable by previous learning algorithms for recurrent neural networks (RNNs). We identify a weakness of LSTM networks processing continual input streams that are not a priori segmented into subsequences with explicitly marked ends at which the network's internal state could be reset. Without resets, the state may grow indefinitely and eventually cause the network to break down. Our remedy is a novel, adaptive "forget gate" that enables an LSTM cell to learn to reset itself at appropriate times, thus releasing internal resources. We review illustrative benchmark problems on which standard LSTM outperforms other RNN algorithms. All algorithms (including LSTM) fail to solve continual versions of these problems. LSTM with forget gates, however, easily solves them, and in an elegant way.     

RNN / LSTM with modified Adam optimizer in deep learning approach for automobile spare parts demand forecasting

The spare parts demand forecasting is very much essential for the organizations to minimize the cost and prevent the stock outs. The demand of spare parts/ car sales distribution is an important factor in inventory control. The valuation of the demand is challenging as the automobile spare parts/car sales demand are often recurrent. The renowned empirical method adopts historical demand data to create the distribution of lead time demand. Although it works reasonably well when service requirements are relatively low, it has difficulty reaching high target service levels. In this paper, we proposed Recurrent Neural Networks/ Long-Short Term Memory (RNN / LSTM) with modified Adam optimizer to predict the demand for spare parts. In this LSTM, weight vectors are generated respectively. These weights are optimized using the Modified-Adam algorithm. The accuracy of the forecast and the performance of the inventory are considered in the experimental result. Experimental results confirm that RNN / LSTM with a Modified-Adam works well with minimal error compared to other existing methods. We conclude that the proposed RNN/LSTM with Modified-Adam algorithm is well suited for the prediction of automobile spare parts.