Particulate nitrate,a key component of fine particles,forms through the intricate gas-to-particle conversion process.This process is regulated by the gas-to-particle conversion coefficient of nitrate(ε(NO_(3)^(-))).T...Particulate nitrate,a key component of fine particles,forms through the intricate gas-to-particle conversion process.This process is regulated by the gas-to-particle conversion coefficient of nitrate(ε(NO_(3)^(-))).The mechanism betweenε(NO_(3)^(-))and its drivers is highly complex and nonlinear,and can be characterized by machine learning methods.However,conventional machine learning often yields results that lack clear physical meaning and may even contradict established physical/chemical mechanisms due to the influence of ambient factors.It urgently needs an alternative approach that possesses transparent physical interpretations and provides deeper insights into the impact ofε(NO_(3)^(-)).Here we introduce a supervised machine learning approachdthe multilevel nested random forest guided by theory approaches.Our approach robustly identifies NH4 t,SO_(4)^(2-),and temperature as pivotal drivers forε(NO_(3)^(-)).Notably,substantial disparities exist between the outcomes of traditional random forest analysis and the anticipated actual results.Furthermore,our approach underscores the significance of NH4 t during both daytime(30%)and nighttime(40%)periods,while appropriately downplaying the influence of some less relevant drivers in comparison to conventional random forest analysis.This research underscores the transformative potential of integrating domain knowledge with machine learning in atmospheric studies.展开更多
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...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.展开更多
基金supported by the National Natural Science Foundation of China(42077191)the National Key Research and Development Program of China(2022YFC3703400)+1 种基金the Blue Sky Foundation,Tianjin Science and Technology Plan Project(18PTZWHZ00120)Fundamental Research Funds for the Central Universities(63213072 and 63213074).
文摘Particulate nitrate,a key component of fine particles,forms through the intricate gas-to-particle conversion process.This process is regulated by the gas-to-particle conversion coefficient of nitrate(ε(NO_(3)^(-))).The mechanism betweenε(NO_(3)^(-))and its drivers is highly complex and nonlinear,and can be characterized by machine learning methods.However,conventional machine learning often yields results that lack clear physical meaning and may even contradict established physical/chemical mechanisms due to the influence of ambient factors.It urgently needs an alternative approach that possesses transparent physical interpretations and provides deeper insights into the impact ofε(NO_(3)^(-)).Here we introduce a supervised machine learning approachdthe multilevel nested random forest guided by theory approaches.Our approach robustly identifies NH4 t,SO_(4)^(2-),and temperature as pivotal drivers forε(NO_(3)^(-)).Notably,substantial disparities exist between the outcomes of traditional random forest analysis and the anticipated actual results.Furthermore,our approach underscores the significance of NH4 t during both daytime(30%)and nighttime(40%)periods,while appropriately downplaying the influence of some less relevant drivers in comparison to conventional random forest analysis.This research underscores the transformative potential of integrating domain knowledge with machine learning in atmospheric studies.
基金This work is supported by the National Natural Science Funds of China(Project No.U19B2036).
文摘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.
文摘车辆目标检测是自动驾驶的重要环节,现有的车辆目标检测算法在特征提取方面没有充分考虑卷积神经网络(convolutional neural network,CNN)和Transformer各自的优缺点,一定程度上限制了网络的整体性能。提出了一种由CNN和Transformer组成的双分支特征聚合网络。在编码阶段,基于CNN和Transformer各自的优势,构建了双分支主干网络来提取原始图像的特征信息;通过设计的多级别空间注意力模块和双支路特征聚合模块,使两个分支间的特征信息相互引导学习;通过构建的双分支注意力模块来进一步减少深层神经网络中特征信息的丢失。在实验部分通过消融实验和对比实验进一步验证了所提算法的有效性,其相比主流的目标检测算法,在mAP(mean average precision)指标上提升了约3.5%。
