高效合理地决策未来月度计划期的铁路运用车分配方案对满足货运市场需求和节约空车调配成本具有重要意义。针对运用车分配方案所具备的时空特性,提出一种基于卷积长短期记忆神经网络并融合物理引导思想的铁路运用车分配预测方法。首先,...高效合理地决策未来月度计划期的铁路运用车分配方案对满足货运市场需求和节约空车调配成本具有重要意义。针对运用车分配方案所具备的时空特性,提出一种基于卷积长短期记忆神经网络并融合物理引导思想的铁路运用车分配预测方法。首先,通过主要影响因素识别,筛选出货运工作量类、货运组织水平类、货车运用属性类3类主要影响指标,并通过滑动窗口划分方法构造为模型输入。随后,构建基于PG-ConvLSTM(convolutional long short-term memory,ConvLSTM)网络的铁路运用车分配预测模型。模型以卷积长短期记忆神经网络作为主体框架,基于运用车分配的先验规律设计物理不一致项构造物理引导损失函数,并使用Hyperband算法优化模型网络层数与物理不一致项权重2项超参数。最后,使用MAE、RMSE与MAPE作为评价指标,并选用BP、CNN-LSTM、CNN-GRU与ConvLSTM网络模型作为对比,基于运用车分配实际数据进行实例分析。结果表明,PG-ConvLSTM模型评价指标MAE为0.0028,RMSE为0.0034,MAPE为7.22%,相比其他神经网络模型均为最优。PG-ConvLSTM模型得益于时空关联特征的同步提取机制,有效避免时空特征经卷积后再输入循环神经网络而丢失关键信息,从而具备更优的预测性能。物理引导损失函数对预测精度提升也具有积极作用。PG-ConvLSTM模型能够高效且准确地对运用车分配方案进行预测,可为实际运营中月度计划期运用车分配方案的制定提供参考。展开更多
Spatiotemporal forecasting of surface soil moisture(SSM)is recognized as a critical scientific issue in precision agricultural irrigation,regional drought monitoring,and early warning systems for extreme precipitation...Spatiotemporal forecasting of surface soil moisture(SSM)is recognized as a critical scientific issue in precision agricultural irrigation,regional drought monitoring,and early warning systems for extreme precipitation.However,long-term forecasting continues to pose formidable challenges because of the complexity observed across both the spatial and temporal scales.In this study,we used a daily SSM dataset at a 0.05°×0.05°spatial resolution over the Qilian Mountains,China and proposed a hybrid Convolutional Long Short-Term Memory(ConvLSTM)-Nudging model,which combined deep neural networks with data assimilation to increase the accuracy of long-term SSM forecasting.We trained and evaluated the SSM predictive performance of four models(Convolutional Neural Network(CNN),Long Short-Term Memory(LSTM),ConvLSTM,and ConvLSTM with Squeeze-and-Excitation(SE)attention mechanism(ConvLSTM-SE))in both short-term and long-term scenarios.The results showed that all the models perform well under short-term predictions,but the accuracy decrease substantially in long-term predictions.Therefore,we integrated Nudging technique during the long-term prediction phase to assimilate observational information and rectify model biases.Comprehensive evaluations demonstrate that Nudging significantly improves all the models,with ConvLSTM-Nudging achieving the best performance under the 200-d forecasting scenario.Relative to those of the best-performing ConvLSTM model for long-term forecasts,when observation noiseδ=0.00 and observation fraction obs=50.0%,the coefficient of determination(R2)of ConvLSTM-Nudging increases by approximately 82.1%,while its mean absolute error(MAE)and root mean squared error(RMSE)decrease by approximately 84.8%and 77.3%,respectively;the average Pearson correlation coefficient(r)improves by approximately 23.6%,and Bias is reduced by 98.1%.These results demonstrated that although pure deep learning models achieve high accuracy in the short-term predictions,they are prone to error accumulation and systematic drift in long-term autoregressive predictions.Integrating data assimilation with deep learning and continuously correcting the state through observation can effectively suppress long-term biases,thereby achieving robust long-term SSM forecasting.展开更多
Photonic platforms are gradually emerging as a promising option to encounter the ever-growing demand for artificial intelligence,among which photonic time-delay reservoir computing(TDRC)is widely anticipated.While suc...Photonic platforms are gradually emerging as a promising option to encounter the ever-growing demand for artificial intelligence,among which photonic time-delay reservoir computing(TDRC)is widely anticipated.While such a computing paradigm can only employ a single photonic device as the nonlinear node for data processing,the performance highly relies on the fading memory provided by the delay feedback loop(FL),which sets a restriction on the extensibility of physical implementation,especially for highly integrated chips.Here,we present a simplified photonic scheme for more flexible parameter configurations leveraging the designed quasi-convolution coding(QC),which completely gets rid of the dependence on FL.Unlike delay-based TDRC,encoded data in QC-based RC(QRC)enables temporal feature extraction,facilitating augmented memory capabilities.Thus,our proposed QRC is enabled to deal with time-related tasks or sequential data without the implementation of FL.Furthermore,we can implement this hardware with a low-power,easily integrable vertical-cavity surface-emitting laser for high-performance parallel processing.We illustrate the concept validation through simulation and experimental comparison of QRC and TDRC,wherein the simpler-structured QRC outperforms across various benchmark tasks.Our results may underscore an auspicious solution for the hardware implementation of deep neural networks.展开更多
BACKGROUND Fear-related disorders,such as post-traumatic stress disorder(PTSD),significantly impact patients and families.Exposure therapy is a common treatment,but imp-roving its effectiveness remains a key challenge...BACKGROUND Fear-related disorders,such as post-traumatic stress disorder(PTSD),significantly impact patients and families.Exposure therapy is a common treatment,but imp-roving its effectiveness remains a key challenge.Fear conditioning and extinction in animal models offer insights into its mechanisms.Our previous research indi-cates that DNA methyltransferases play a role in fear memory renewal.AIM To investigate the role of DNA methylation in the extinction of fear memory,with the goal of identifying potential strategies to enhance the efficacy of exposure therapy for fear-related disorders.METHODS This study investigated the role of DNA methylation in fear memory extinction in mice.DNA methylation was manipulated using N-phthalyl-L-tryptophan(RG108)to reduce methylation and L-methionine injections to enhance it.Neuronal activity,and dendritic spine density was measured following extinction training.RESULTS RG108 suppressed extinction,reduced spine density,and inhibited neuronal activity.Methionine injections facilitated extinction.CONCLUSION DNA methylation is crucial for fear memory extinction.Enhancing methylation may improve the efficacy of exposure therapy,offering a potential strategy to treat fear-related disorders.展开更多
Background:Over the past few decades,a threefold increase in obesity and type 2 diabetes(T2D)has placed a heavy burden on the health-care system and society.Previous studies have shown correlations between obesity,T2D...Background:Over the past few decades,a threefold increase in obesity and type 2 diabetes(T2D)has placed a heavy burden on the health-care system and society.Previous studies have shown correlations between obesity,T2D,and neurodegenera-tive diseases,including dementia.It is imperative to further understand the relation-ship between obesity,T2D,and cognitive deficits.Methods:This investigation tested and evaluated the cognitive impact of obesity and T2D induced by high-fat diet(HFD)and the effect of the host genetic background on the severity of cognitive decline caused by obesity and T2D in collaborative cross(CC)mice.The CC mice are a genetically diverse panel derived from eight inbred strains.Results:Our findings demonstrated significant variations in the recorded phenotypes across different CC lines compared to the reference mouse line,C57BL/6J.CC037 line exhibited a substantial increase in body weight on HFD,whereas line CC005 ex-hibited differing responses based on sex.Glucose tolerance tests revealed significant variations,with some lines like CC005 showing a marked increase in area under the curve(AUC)values on HFD.Organ weights,including brain,spleen,liver,and kidney,varied significantly among the lines and sexes in response to HFD.Behavioral tests using the Morris water maze indicated that cognitive performance was differentially affected by diet and genetic background.Conclusions:Our study establishes a foundation for future quantitative trait loci map-ping using CC lines and identifying genes underlying the comorbidity of Alzheimer's disease(AD),caused by obesity and T2D.The genetic components may offer new tools for early prediction and prevention.展开更多
This study investigates the application of Learnable Memory Vision Transformers(LMViT)for detecting metal surface flaws,comparing their performance with traditional CNNs,specifically ResNet18 and ResNet50,as well as o...This study investigates the application of Learnable Memory Vision Transformers(LMViT)for detecting metal surface flaws,comparing their performance with traditional CNNs,specifically ResNet18 and ResNet50,as well as other transformer-based models including Token to Token ViT,ViT withoutmemory,and Parallel ViT.Leveraging awidely-used steel surface defect dataset,the research applies data augmentation and t-distributed stochastic neighbor embedding(t-SNE)to enhance feature extraction and understanding.These techniques mitigated overfitting,stabilized training,and improved generalization capabilities.The LMViT model achieved a test accuracy of 97.22%,significantly outperforming ResNet18(88.89%)and ResNet50(88.90%),aswell as the Token to TokenViT(88.46%),ViT without memory(87.18),and Parallel ViT(91.03%).Furthermore,LMViT exhibited superior training and validation performance,attaining a validation accuracy of 98.2%compared to 91.0%for ResNet 18,96.0%for ResNet50,and 89.12%,87.51%,and 91.21%for Token to Token ViT,ViT without memory,and Parallel ViT,respectively.The findings highlight the LMViT’s ability to capture long-range dependencies in images,an areawhere CNNs struggle due to their reliance on local receptive fields and hierarchical feature extraction.The additional transformer-based models also demonstrate improved performance in capturing complex features over CNNs,with LMViT excelling particularly at detecting subtle and complex defects,which is critical for maintaining product quality and operational efficiency in industrial applications.For instance,the LMViT model successfully identified fine scratches and minor surface irregularities that CNNs often misclassify.This study not only demonstrates LMViT’s potential for real-world defect detection but also underscores the promise of other transformer-based architectures like Token to Token ViT,ViT without memory,and Parallel ViT in industrial scenarios where complex spatial relationships are key.Future research may focus on enhancing LMViT’s computational efficiency for deployment in real-time quality control systems.展开更多
本文应用ConvLSTM (Convolutional Long Short-Term Memory)模型解决了空气污染物浓度短时预测的问题。首先基于卷积神经网络和长短期记忆网络对ConvLSTM模型的构建方法进行了探讨,深入剖析了模型的基本构成与结构特性,然后通过实验实...本文应用ConvLSTM (Convolutional Long Short-Term Memory)模型解决了空气污染物浓度短时预测的问题。首先基于卷积神经网络和长短期记忆网络对ConvLSTM模型的构建方法进行了探讨,深入剖析了模型的基本构成与结构特性,然后通过实验实例详细展示了该模型在空气污染物浓度预测领域的应用过程,包括数据预处理、模型训练、预测结果分析等。实验结果表明,ConvLSTM模型用于空气污染物浓度短时预测的精度较高(始终保持在0.1%以内),同时也表明模型预测精度与时间序列并非总是正相关。当时间步长在某个值(本文实验中时间步长为10)附近时,模型预测精度较高。本研究可为其他具有类似时空特征数据序列的预测问题提供参考。展开更多
针对现有雷达回波外推模型存在长时序回波外推模糊失真和强回波预测准确率较低等问题,利用安徽2016年5-9月的多普勒雷达组合反射率拼图数据,设计了一种基于自注意力和稠密卷积改进卷积长短期记忆(convolutional long short-term memory,...针对现有雷达回波外推模型存在长时序回波外推模糊失真和强回波预测准确率较低等问题,利用安徽2016年5-9月的多普勒雷达组合反射率拼图数据,设计了一种基于自注意力和稠密卷积改进卷积长短期记忆(convolutional long short-term memory,ConvLSTM)网络的雷达回波外推方法。模型以ConvLSTM为基础,在每个单元结构以及编解码器中间融入自注意力机制,强化模型对于特征长时空间依赖的提取能力,同时用稠密连接卷积代替普通卷积,提高模型的特征重用能力。实验利用过去1 h雷达回波图像预测未来2 h雷达回波图像,并与改进前的ConvLSTM进行对比证明了提出的模型能够提高雷达回波外推的准确率。展开更多
文摘针对在多传感器下变转速且带有不同程度噪声的工况下故障特征被淹没的问题,提出一种基于改进卷积长短时记忆网络(Convolutional LSTM, ConvLSTM)的故障诊断方法:首先将多个传感器采集的一维振动信号切分为二维矩阵序列;再利用由改进ConvLSTM单元构成的特征提取层提取信号内的时间特征和空间特征,改进ConvLSTM单元是将传统ConvLSTM单元输入门中的普通卷积换成膨胀卷积,在相同的卷积核其有更大的感受野读取输入信息;最后通过由卷积层和全局平均池化(Global Average Pooling,GAP)构造的分类输出层得到诊断结果。试验使用CWRU滚动轴承数据集和XJTU-SY滚动轴承数据集进行验证。试验结果表明,与其他对比模型相比,改进ConvLSTM模型在变转速且带有不同程度噪声下达到较高的精确率并且受样本量的影响更小。
文摘高效合理地决策未来月度计划期的铁路运用车分配方案对满足货运市场需求和节约空车调配成本具有重要意义。针对运用车分配方案所具备的时空特性,提出一种基于卷积长短期记忆神经网络并融合物理引导思想的铁路运用车分配预测方法。首先,通过主要影响因素识别,筛选出货运工作量类、货运组织水平类、货车运用属性类3类主要影响指标,并通过滑动窗口划分方法构造为模型输入。随后,构建基于PG-ConvLSTM(convolutional long short-term memory,ConvLSTM)网络的铁路运用车分配预测模型。模型以卷积长短期记忆神经网络作为主体框架,基于运用车分配的先验规律设计物理不一致项构造物理引导损失函数,并使用Hyperband算法优化模型网络层数与物理不一致项权重2项超参数。最后,使用MAE、RMSE与MAPE作为评价指标,并选用BP、CNN-LSTM、CNN-GRU与ConvLSTM网络模型作为对比,基于运用车分配实际数据进行实例分析。结果表明,PG-ConvLSTM模型评价指标MAE为0.0028,RMSE为0.0034,MAPE为7.22%,相比其他神经网络模型均为最优。PG-ConvLSTM模型得益于时空关联特征的同步提取机制,有效避免时空特征经卷积后再输入循环神经网络而丢失关键信息,从而具备更优的预测性能。物理引导损失函数对预测精度提升也具有积极作用。PG-ConvLSTM模型能够高效且准确地对运用车分配方案进行预测,可为实际运营中月度计划期运用车分配方案的制定提供参考。
基金funded by the National Natural Science Foundation of China(42461053)the Department of Education of Gansu Province:Higher Education Innovation Fund Project(2023B-064)+1 种基金the Youth Doctoral Fund Project(2024QB-014)the Natural Science Foundation of Gansu Province(25JRRA012).
文摘Spatiotemporal forecasting of surface soil moisture(SSM)is recognized as a critical scientific issue in precision agricultural irrigation,regional drought monitoring,and early warning systems for extreme precipitation.However,long-term forecasting continues to pose formidable challenges because of the complexity observed across both the spatial and temporal scales.In this study,we used a daily SSM dataset at a 0.05°×0.05°spatial resolution over the Qilian Mountains,China and proposed a hybrid Convolutional Long Short-Term Memory(ConvLSTM)-Nudging model,which combined deep neural networks with data assimilation to increase the accuracy of long-term SSM forecasting.We trained and evaluated the SSM predictive performance of four models(Convolutional Neural Network(CNN),Long Short-Term Memory(LSTM),ConvLSTM,and ConvLSTM with Squeeze-and-Excitation(SE)attention mechanism(ConvLSTM-SE))in both short-term and long-term scenarios.The results showed that all the models perform well under short-term predictions,but the accuracy decrease substantially in long-term predictions.Therefore,we integrated Nudging technique during the long-term prediction phase to assimilate observational information and rectify model biases.Comprehensive evaluations demonstrate that Nudging significantly improves all the models,with ConvLSTM-Nudging achieving the best performance under the 200-d forecasting scenario.Relative to those of the best-performing ConvLSTM model for long-term forecasts,when observation noiseδ=0.00 and observation fraction obs=50.0%,the coefficient of determination(R2)of ConvLSTM-Nudging increases by approximately 82.1%,while its mean absolute error(MAE)and root mean squared error(RMSE)decrease by approximately 84.8%and 77.3%,respectively;the average Pearson correlation coefficient(r)improves by approximately 23.6%,and Bias is reduced by 98.1%.These results demonstrated that although pure deep learning models achieve high accuracy in the short-term predictions,they are prone to error accumulation and systematic drift in long-term autoregressive predictions.Integrating data assimilation with deep learning and continuously correcting the state through observation can effectively suppress long-term biases,thereby achieving robust long-term SSM forecasting.
基金National Natural Science Foundation of China(62171305,62405206,62004135,62001317,62111530301)Natural Science Foundation of Jiangsu Province(BK20240778,BK20241917)+3 种基金State Key Laboratory of Advanced Optical Communication Systems and Networks,China(2023GZKF08)China Postdoctoral Science Foundation(2024M752314)Postdoctoral Fellowship Program of CPSF(GZC20231883)Innovative and Entrepreneurial Talent Program of Jiangsu Province(JSSCRC2021527).
文摘Photonic platforms are gradually emerging as a promising option to encounter the ever-growing demand for artificial intelligence,among which photonic time-delay reservoir computing(TDRC)is widely anticipated.While such a computing paradigm can only employ a single photonic device as the nonlinear node for data processing,the performance highly relies on the fading memory provided by the delay feedback loop(FL),which sets a restriction on the extensibility of physical implementation,especially for highly integrated chips.Here,we present a simplified photonic scheme for more flexible parameter configurations leveraging the designed quasi-convolution coding(QC),which completely gets rid of the dependence on FL.Unlike delay-based TDRC,encoded data in QC-based RC(QRC)enables temporal feature extraction,facilitating augmented memory capabilities.Thus,our proposed QRC is enabled to deal with time-related tasks or sequential data without the implementation of FL.Furthermore,we can implement this hardware with a low-power,easily integrable vertical-cavity surface-emitting laser for high-performance parallel processing.We illustrate the concept validation through simulation and experimental comparison of QRC and TDRC,wherein the simpler-structured QRC outperforms across various benchmark tasks.Our results may underscore an auspicious solution for the hardware implementation of deep neural networks.
基金Supported by National Natural Science Foundation of China,No.82360231Yunnan Basic Research Program General Project,No.202401AT070075+1 种基金Dali Basic Research Program Key Project,No.202301A020021Youth Special Project for Basic Research of Local Universities in Yunnan Province,No.202301BA070001-127.
文摘BACKGROUND Fear-related disorders,such as post-traumatic stress disorder(PTSD),significantly impact patients and families.Exposure therapy is a common treatment,but imp-roving its effectiveness remains a key challenge.Fear conditioning and extinction in animal models offer insights into its mechanisms.Our previous research indi-cates that DNA methyltransferases play a role in fear memory renewal.AIM To investigate the role of DNA methylation in the extinction of fear memory,with the goal of identifying potential strategies to enhance the efficacy of exposure therapy for fear-related disorders.METHODS This study investigated the role of DNA methylation in fear memory extinction in mice.DNA methylation was manipulated using N-phthalyl-L-tryptophan(RG108)to reduce methylation and L-methionine injections to enhance it.Neuronal activity,and dendritic spine density was measured following extinction training.RESULTS RG108 suppressed extinction,reduced spine density,and inhibited neuronal activity.Methionine injections facilitated extinction.CONCLUSION DNA methylation is crucial for fear memory extinction.Enhancing methylation may improve the efficacy of exposure therapy,offering a potential strategy to treat fear-related disorders.
文摘Background:Over the past few decades,a threefold increase in obesity and type 2 diabetes(T2D)has placed a heavy burden on the health-care system and society.Previous studies have shown correlations between obesity,T2D,and neurodegenera-tive diseases,including dementia.It is imperative to further understand the relation-ship between obesity,T2D,and cognitive deficits.Methods:This investigation tested and evaluated the cognitive impact of obesity and T2D induced by high-fat diet(HFD)and the effect of the host genetic background on the severity of cognitive decline caused by obesity and T2D in collaborative cross(CC)mice.The CC mice are a genetically diverse panel derived from eight inbred strains.Results:Our findings demonstrated significant variations in the recorded phenotypes across different CC lines compared to the reference mouse line,C57BL/6J.CC037 line exhibited a substantial increase in body weight on HFD,whereas line CC005 ex-hibited differing responses based on sex.Glucose tolerance tests revealed significant variations,with some lines like CC005 showing a marked increase in area under the curve(AUC)values on HFD.Organ weights,including brain,spleen,liver,and kidney,varied significantly among the lines and sexes in response to HFD.Behavioral tests using the Morris water maze indicated that cognitive performance was differentially affected by diet and genetic background.Conclusions:Our study establishes a foundation for future quantitative trait loci map-ping using CC lines and identifying genes underlying the comorbidity of Alzheimer's disease(AD),caused by obesity and T2D.The genetic components may offer new tools for early prediction and prevention.
基金funded by Woosong University Academic Research 2024.
文摘This study investigates the application of Learnable Memory Vision Transformers(LMViT)for detecting metal surface flaws,comparing their performance with traditional CNNs,specifically ResNet18 and ResNet50,as well as other transformer-based models including Token to Token ViT,ViT withoutmemory,and Parallel ViT.Leveraging awidely-used steel surface defect dataset,the research applies data augmentation and t-distributed stochastic neighbor embedding(t-SNE)to enhance feature extraction and understanding.These techniques mitigated overfitting,stabilized training,and improved generalization capabilities.The LMViT model achieved a test accuracy of 97.22%,significantly outperforming ResNet18(88.89%)and ResNet50(88.90%),aswell as the Token to TokenViT(88.46%),ViT without memory(87.18),and Parallel ViT(91.03%).Furthermore,LMViT exhibited superior training and validation performance,attaining a validation accuracy of 98.2%compared to 91.0%for ResNet 18,96.0%for ResNet50,and 89.12%,87.51%,and 91.21%for Token to Token ViT,ViT without memory,and Parallel ViT,respectively.The findings highlight the LMViT’s ability to capture long-range dependencies in images,an areawhere CNNs struggle due to their reliance on local receptive fields and hierarchical feature extraction.The additional transformer-based models also demonstrate improved performance in capturing complex features over CNNs,with LMViT excelling particularly at detecting subtle and complex defects,which is critical for maintaining product quality and operational efficiency in industrial applications.For instance,the LMViT model successfully identified fine scratches and minor surface irregularities that CNNs often misclassify.This study not only demonstrates LMViT’s potential for real-world defect detection but also underscores the promise of other transformer-based architectures like Token to Token ViT,ViT without memory,and Parallel ViT in industrial scenarios where complex spatial relationships are key.Future research may focus on enhancing LMViT’s computational efficiency for deployment in real-time quality control systems.
文摘本文应用ConvLSTM (Convolutional Long Short-Term Memory)模型解决了空气污染物浓度短时预测的问题。首先基于卷积神经网络和长短期记忆网络对ConvLSTM模型的构建方法进行了探讨,深入剖析了模型的基本构成与结构特性,然后通过实验实例详细展示了该模型在空气污染物浓度预测领域的应用过程,包括数据预处理、模型训练、预测结果分析等。实验结果表明,ConvLSTM模型用于空气污染物浓度短时预测的精度较高(始终保持在0.1%以内),同时也表明模型预测精度与时间序列并非总是正相关。当时间步长在某个值(本文实验中时间步长为10)附近时,模型预测精度较高。本研究可为其他具有类似时空特征数据序列的预测问题提供参考。
