海关商品HS编码分类是企业和个人进出口贸易的重要国际程序。HS编码分类可以看作是一个文本分类问题,即给定一段商品的描述,确定商品由HS编码表示的所属类别。然而,该任务比一般的文本分类任务更具挑战性,原因是商品描述文本具有特定的...海关商品HS编码分类是企业和个人进出口贸易的重要国际程序。HS编码分类可以看作是一个文本分类问题,即给定一段商品的描述,确定商品由HS编码表示的所属类别。然而,该任务比一般的文本分类任务更具挑战性,原因是商品描述文本具有特定的层次结构,同时商品描述文本展现出了两个层次上的序列特征,并且商品描述文本还存在关键信息分散且描述形式多样的特点。现有的文本分类方法无法综合考虑以上因素来捕获商品描述文本中的关键信息。对此,文中提出了一种融合文本序列和图信息的神经网络(Text Sequence and Graph Information combination Neural Network,TSGINN)模型,用于解决海关商品HS编码分类问题。TSGINN将HS编码分类问题定义为基于词共现网络的子图分类问题,通过图注意力网络建模非连续词之间的关联关系,同时利用分层的长短期记忆网络结合商品文本层次结构捕获多层次的序列信息。在真实海关商品数据集上进行了实验,结果表明TSGINN模型的HS编码分类效果优于其他分类方法。展开更多
Present study provides a simple analytical formula,the“Klingel-like formula”or“Pascal’s Formula”that can be used as a reference to test some results of existing railway codes and specifically those using rigid co...Present study provides a simple analytical formula,the“Klingel-like formula”or“Pascal’s Formula”that can be used as a reference to test some results of existing railway codes and specifically those using rigid contact.It develops properly the 3D Newton-Euler equations governing the 6 degrees of freedom(DoF)of unsuspended loaded wheelsets in case of zero wheel-rail friction and constant conicity.Thus,by solving numerically these equations,we got pendulum like harmonic oscillations of which the calculated angular frequency is used for assessing the accuracy of the proposed formula so that it can in turn be used as a fast practical target for testing multi-body system(MBS)railway codes.Due to the harmonic property of these pendulum-like oscillations,the squareω2 of their angular frequency can be made in the form of a ratio K/M where K depends on the wheelset geometry and load and M on its inertia.Information on K and M are useful to understand wheelsets behavior.The analytical formula is derived from the first order writing of full trigonometric Newton-Euler equations by setting zero elastic wheel-rail penetration and by assuming small displacements.Full trigonometric equations are numerically solved to assess that the formula providesω2 inside a 1%accuracy for usual wheelsets dimensions.By decreasing the conicity down to 1×10^(−4) rad,the relative formula accuracy is under 3×10^(−5).In order to test the formula reliability for rigid contact formulations,the stiffness of elastic contacts can be increased up to practical rigidity(Hertz stiffness×1000).展开更多
Bit-Interleaved Coded Modulation with Iterative Decoding (BICM-ID) is a bandwidth ef- ficient transmission, where the bit error rate is reduced through the iterative information exchange between the inner demapper and...Bit-Interleaved Coded Modulation with Iterative Decoding (BICM-ID) is a bandwidth ef- ficient transmission, where the bit error rate is reduced through the iterative information exchange between the inner demapper and the outer decoder. The choice of the symbol mapping is the crucial design parameter. This paper indicates that the Harmonic Mean of the Minimum Squared Euclidean (HMMSE) distance is the best criterion for the mapping design. Based on the design criterion of the HMMSE distance, a new search algorithm to find the optimized labeling maps for BICM-ID system is proposed. Numerical results and performance comparison show that the new labeling search method has a low complexity and outperforms other labeling schemes using other design criterion in BICM-ID system, therefore it is an optimized labeling method.展开更多
文摘海关商品HS编码分类是企业和个人进出口贸易的重要国际程序。HS编码分类可以看作是一个文本分类问题,即给定一段商品的描述,确定商品由HS编码表示的所属类别。然而,该任务比一般的文本分类任务更具挑战性,原因是商品描述文本具有特定的层次结构,同时商品描述文本展现出了两个层次上的序列特征,并且商品描述文本还存在关键信息分散且描述形式多样的特点。现有的文本分类方法无法综合考虑以上因素来捕获商品描述文本中的关键信息。对此,文中提出了一种融合文本序列和图信息的神经网络(Text Sequence and Graph Information combination Neural Network,TSGINN)模型,用于解决海关商品HS编码分类问题。TSGINN将HS编码分类问题定义为基于词共现网络的子图分类问题,通过图注意力网络建模非连续词之间的关联关系,同时利用分层的长短期记忆网络结合商品文本层次结构捕获多层次的序列信息。在真实海关商品数据集上进行了实验,结果表明TSGINN模型的HS编码分类效果优于其他分类方法。
文摘Present study provides a simple analytical formula,the“Klingel-like formula”or“Pascal’s Formula”that can be used as a reference to test some results of existing railway codes and specifically those using rigid contact.It develops properly the 3D Newton-Euler equations governing the 6 degrees of freedom(DoF)of unsuspended loaded wheelsets in case of zero wheel-rail friction and constant conicity.Thus,by solving numerically these equations,we got pendulum like harmonic oscillations of which the calculated angular frequency is used for assessing the accuracy of the proposed formula so that it can in turn be used as a fast practical target for testing multi-body system(MBS)railway codes.Due to the harmonic property of these pendulum-like oscillations,the squareω2 of their angular frequency can be made in the form of a ratio K/M where K depends on the wheelset geometry and load and M on its inertia.Information on K and M are useful to understand wheelsets behavior.The analytical formula is derived from the first order writing of full trigonometric Newton-Euler equations by setting zero elastic wheel-rail penetration and by assuming small displacements.Full trigonometric equations are numerically solved to assess that the formula providesω2 inside a 1%accuracy for usual wheelsets dimensions.By decreasing the conicity down to 1×10^(−4) rad,the relative formula accuracy is under 3×10^(−5).In order to test the formula reliability for rigid contact formulations,the stiffness of elastic contacts can be increased up to practical rigidity(Hertz stiffness×1000).
基金Supported by the National Natural Science Foundation of China (No.60472104).
文摘Bit-Interleaved Coded Modulation with Iterative Decoding (BICM-ID) is a bandwidth ef- ficient transmission, where the bit error rate is reduced through the iterative information exchange between the inner demapper and the outer decoder. The choice of the symbol mapping is the crucial design parameter. This paper indicates that the Harmonic Mean of the Minimum Squared Euclidean (HMMSE) distance is the best criterion for the mapping design. Based on the design criterion of the HMMSE distance, a new search algorithm to find the optimized labeling maps for BICM-ID system is proposed. Numerical results and performance comparison show that the new labeling search method has a low complexity and outperforms other labeling schemes using other design criterion in BICM-ID system, therefore it is an optimized labeling method.
