Motor imagery is the mental representation of an action without overt movement or muscle activation. However, the effects of motor imagery on stroke-induced hand dysfunction and brain neural networks are still unknown...Motor imagery is the mental representation of an action without overt movement or muscle activation. However, the effects of motor imagery on stroke-induced hand dysfunction and brain neural networks are still unknown. We conducted a randomized controlled trial in the China Rehabilitation Research Center. Twenty stroke patients, including 13 males and 7 females, 32–51 years old, were recruited and randomly assigned to the traditional rehabilitation treatment group（PP group, n = 10） or the motor imagery training combined with traditional rehabilitation treatment group（MP group, n = 10）. All patients received rehabilitation training once a day, 45 minutes per session, five times per week, for 4 consecutive weeks. In the MP group, motor imagery training was performed for 45 minutes after traditional rehabilitation training, daily. Action Research Arm Test and the Fugl-Meyer Assessment of the upper extremity were used to evaluate hand functions before and after treatment. Transcranial magnetic stimulation was used to analyze motor evoked potentials in the affected extremity. Diffusion tensor imaging was used to assess changes in brain neural networks. Compared with the PP group, the MP group showed better recovery of hand function, higher amplitude of the motor evoked potential in the abductor pollicis brevis, greater fractional anisotropy of the right dorsal pathway, and an increase in the fractional anisotropy of the bilateral dorsal pathway. Our findings indicate that 4 weeks of motor imagery training combined with traditional rehabilitation treatment improves hand function in stroke patients by enhancing the dorsal pathway. This trial has been registered with the Chinese Clinical Trial Registry（registration number： Chi CTR-OCH-12002238）.展开更多
In this paper,we introduce a type of tensor neural network.For the first time,we propose its numerical integration scheme and prove the computational complexity to be the polynomial scale of the dimension.Based on the...In this paper,we introduce a type of tensor neural network.For the first time,we propose its numerical integration scheme and prove the computational complexity to be the polynomial scale of the dimension.Based on the tensor product structure,we develop an efficient numerical integration method by using fixed quadrature points for the functions of the tensor neural network.The corresponding machine learning method is also introduced for solving high-dimensional problems.Some numerical examples are also provided to validate the theoretical results and the numerical algorithm.展开更多
The present study aims to estimate the in vivo anisotropic conductivities of the White Matter (WM) tissues by means of Magnetic Resonance Electrical Impedance Tomography (MREIT) technique. The realistic anisotropic vo...The present study aims to estimate the in vivo anisotropic conductivities of the White Matter (WM) tissues by means of Magnetic Resonance Electrical Impedance Tomography (MREIT) technique. The realistic anisotropic volume conductor model with different conductivity properties (scalp, skull, CSF, gray matter and WM) is constructed based on the Diffusion Tensor Magnetic Resonance Imaging (DT- MRI) from a healthy human subject. The Radius Basic Function (RBF)-MREIT algorithm of using only one magnetic flux density component was applied to evaluate the eigenvalues of the anisotropic WM with target values set according to the DT-MRI data based on the Wolter’s model, which is more physiologically reliable. The numerical simulations study performed on the five-layer realistic human head model showed that the conductivity reconstruction method had higher accuracy and better robustness against noise. The pilot research was used to judge the feasibility, meaningfulness and reliability of the MREIT applied on the electrical impedance tomography of the complicated human head tissues including anisotropic characteristics.展开更多
随着人工智能的发展,深度神经网络成为多种模式识别任务中必不可少的工具,由于深度卷积神经网络(CNN)参数量巨大、计算复杂度高,将它部署到计算资源和存储空间受限的边缘计算设备上成为一项挑战。因此,深度网络压缩成为近年来的研究热...随着人工智能的发展,深度神经网络成为多种模式识别任务中必不可少的工具,由于深度卷积神经网络(CNN)参数量巨大、计算复杂度高,将它部署到计算资源和存储空间受限的边缘计算设备上成为一项挑战。因此,深度网络压缩成为近年来的研究热点。低秩分解与向量量化是深度网络压缩中重要的两个研究分支,其核心思想都是通过找到原网络结构的一种紧凑型表达,从而降低网络参数的冗余程度。通过建立联合压缩框架,提出一种基于低秩分解和向量量化的深度网络压缩方法——可量化的张量分解(QTD)。该方法能够在网络低秩结构的基础上实现进一步的量化,从而得到更大的压缩比。在CIFAR-10数据集上对经典ResNet和该方法进行验证的实验结果表明,QTD能够在准确率仅损失1.71个百分点的情况下,将网络参数量压缩至原来的1%。而在大型数据集ImageNet上把所提方法与基于量化的方法PQF(Permute,Quantize,and Fine-tune)、基于低秩分解的方法TDNR(Tucker Decomposition with Nonlinear Response)和基于剪枝的方法CLIP-Q(Compression Learning by In-parallel Pruning-Quantization)进行比较与分析的实验结果表明,QTD能够在相同压缩范围下实现更好的分类准确率。展开更多
Relation classification is a crucial component in many Natural Language Processing(NLP) systems. In this paper, we propose a novel bidirectional recurrent neural network architecture(using Long Short-Term Memory,LSTM,...Relation classification is a crucial component in many Natural Language Processing(NLP) systems. In this paper, we propose a novel bidirectional recurrent neural network architecture(using Long Short-Term Memory,LSTM, cells) for relation classification, with an attention layer for organizing the context information on the word level and a tensor layer for detecting complex connections between two entities. The above two feature extraction operations are based on the LSTM networks and use their outputs. Our model allows end-to-end learning from the raw sentences in the dataset, without trimming or reconstructing them. Experiments on the SemEval-2010 Task 8dataset show that our model outperforms most state-of-the-art methods.展开更多
基金supported by the National Natural Science Foundation of China,No.U1613228a grant from the Sub-Project under National “Twelfth Five-Year” Plan for Science & Technology Support Project in China,No.2011BAI08B11+1 种基金a grant from the Beijing Municipal Science & Technology Commission in China,No.Z161100002616018the Special Fund for Basic Scientific Research Business of Central Public Scientific Research Institutes in China,No.2014CZ-5,2015CZ-30
文摘Motor imagery is the mental representation of an action without overt movement or muscle activation. However, the effects of motor imagery on stroke-induced hand dysfunction and brain neural networks are still unknown. We conducted a randomized controlled trial in the China Rehabilitation Research Center. Twenty stroke patients, including 13 males and 7 females, 32–51 years old, were recruited and randomly assigned to the traditional rehabilitation treatment group（PP group, n = 10） or the motor imagery training combined with traditional rehabilitation treatment group（MP group, n = 10）. All patients received rehabilitation training once a day, 45 minutes per session, five times per week, for 4 consecutive weeks. In the MP group, motor imagery training was performed for 45 minutes after traditional rehabilitation training, daily. Action Research Arm Test and the Fugl-Meyer Assessment of the upper extremity were used to evaluate hand functions before and after treatment. Transcranial magnetic stimulation was used to analyze motor evoked potentials in the affected extremity. Diffusion tensor imaging was used to assess changes in brain neural networks. Compared with the PP group, the MP group showed better recovery of hand function, higher amplitude of the motor evoked potential in the abductor pollicis brevis, greater fractional anisotropy of the right dorsal pathway, and an increase in the fractional anisotropy of the bilateral dorsal pathway. Our findings indicate that 4 weeks of motor imagery training combined with traditional rehabilitation treatment improves hand function in stroke patients by enhancing the dorsal pathway. This trial has been registered with the Chinese Clinical Trial Registry（registration number： Chi CTR-OCH-12002238）.
基金supported in part by the National Key Research and Development Program of China(Grant No.2019YFA0709601)by the National Center for Mathematics and Interdisciplinary Science,CAS.
文摘In this paper,we introduce a type of tensor neural network.For the first time,we propose its numerical integration scheme and prove the computational complexity to be the polynomial scale of the dimension.Based on the tensor product structure,we develop an efficient numerical integration method by using fixed quadrature points for the functions of the tensor neural network.The corresponding machine learning method is also introduced for solving high-dimensional problems.Some numerical examples are also provided to validate the theoretical results and the numerical algorithm.
文摘The present study aims to estimate the in vivo anisotropic conductivities of the White Matter (WM) tissues by means of Magnetic Resonance Electrical Impedance Tomography (MREIT) technique. The realistic anisotropic volume conductor model with different conductivity properties (scalp, skull, CSF, gray matter and WM) is constructed based on the Diffusion Tensor Magnetic Resonance Imaging (DT- MRI) from a healthy human subject. The Radius Basic Function (RBF)-MREIT algorithm of using only one magnetic flux density component was applied to evaluate the eigenvalues of the anisotropic WM with target values set according to the DT-MRI data based on the Wolter’s model, which is more physiologically reliable. The numerical simulations study performed on the five-layer realistic human head model showed that the conductivity reconstruction method had higher accuracy and better robustness against noise. The pilot research was used to judge the feasibility, meaningfulness and reliability of the MREIT applied on the electrical impedance tomography of the complicated human head tissues including anisotropic characteristics.
文摘随着人工智能的发展,深度神经网络成为多种模式识别任务中必不可少的工具,由于深度卷积神经网络(CNN)参数量巨大、计算复杂度高,将它部署到计算资源和存储空间受限的边缘计算设备上成为一项挑战。因此,深度网络压缩成为近年来的研究热点。低秩分解与向量量化是深度网络压缩中重要的两个研究分支,其核心思想都是通过找到原网络结构的一种紧凑型表达,从而降低网络参数的冗余程度。通过建立联合压缩框架,提出一种基于低秩分解和向量量化的深度网络压缩方法——可量化的张量分解(QTD)。该方法能够在网络低秩结构的基础上实现进一步的量化,从而得到更大的压缩比。在CIFAR-10数据集上对经典ResNet和该方法进行验证的实验结果表明,QTD能够在准确率仅损失1.71个百分点的情况下,将网络参数量压缩至原来的1%。而在大型数据集ImageNet上把所提方法与基于量化的方法PQF(Permute,Quantize,and Fine-tune)、基于低秩分解的方法TDNR(Tucker Decomposition with Nonlinear Response)和基于剪枝的方法CLIP-Q(Compression Learning by In-parallel Pruning-Quantization)进行比较与分析的实验结果表明,QTD能够在相同压缩范围下实现更好的分类准确率。
基金supported by the National Natural Science Foundation of China (No. 61572505)ChanXueYan Prospective Project of Jiangsu Province (No. BY201502305)
文摘Relation classification is a crucial component in many Natural Language Processing(NLP) systems. In this paper, we propose a novel bidirectional recurrent neural network architecture(using Long Short-Term Memory,LSTM, cells) for relation classification, with an attention layer for organizing the context information on the word level and a tensor layer for detecting complex connections between two entities. The above two feature extraction operations are based on the LSTM networks and use their outputs. Our model allows end-to-end learning from the raw sentences in the dataset, without trimming or reconstructing them. Experiments on the SemEval-2010 Task 8dataset show that our model outperforms most state-of-the-art methods.
