Accurate handwriting recognition has been a challenging computer vision problem,because static feature analysis of the text pictures is often inade-quate to account for high variance in handwriting styles across peopl...Accurate handwriting recognition has been a challenging computer vision problem,because static feature analysis of the text pictures is often inade-quate to account for high variance in handwriting styles across people and poor image quality of the handwritten text.Recently,by introducing machine learning,especially convolutional neural networks(CNNs),the recognition accuracy of various handwriting patterns is steadily improved.In this paper,a deep CNN model is developed to further improve the recognition rate of the MNIST hand-written digit dataset with a fast-converging rate in training.The proposed model comes with a multi-layer deep arrange structure,including 3 convolution and acti-vation layers for feature extraction and 2 fully connected layers(i.e.,dense layers)for classification.The model’s hyperparameters,such as the batch sizes,kernel sizes,batch normalization,activation function,and learning rate are optimized to enhance the recognition performance.The average classification accuracy of the proposed methodology is found to reach 99.82%on the training dataset and 99.40%on the testing dataset,making it a nearly error-free system for MNIST recognition.展开更多
In this paper, classical and continuous variable (CV) quantum neural network hybrid multi-classifiers are presented using the MNIST dataset. Currently available classifiers can classify only up to two classes. The pro...In this paper, classical and continuous variable (CV) quantum neural network hybrid multi-classifiers are presented using the MNIST dataset. Currently available classifiers can classify only up to two classes. The proposed architecture allows networks to classify classes up to n<sup>m</sup> classes, where n represents cutoff dimension and m the number of qumodes on photonic quantum computers. The combination of cutoff dimension and probability measurement method in the CV model allows a quantum circuit to produce output vectors of size n<sup>m</sup>. They are then interpreted as one-hot encoded labels, padded with n<sup>m</sup> - 10 zeros. The total of seven different classifiers is built using 2, 3, …, 6, and 8-qumodes on photonic quantum computing simulators, based on the binary classifier architecture proposed in “Continuous variable quantum neural networks” [1]. They are composed of a classical feed-forward neural network, a quantum data encoding circuit, and a CV quantum neural network circuit. On a truncated MNIST dataset of 600 samples, a 4-qumode hybrid classifier achieves 100% training accuracy.展开更多
单隐层前馈神经网络是一种高效且结构简单的神经网络,它的一种典型的学习算法就是误差反向传播(error back propagation,BP)算法.这种算法基于最速下降法原理,主要缺点是学习速度过慢.超限学习机(extreme learning machine,ELM)极大地...单隐层前馈神经网络是一种高效且结构简单的神经网络,它的一种典型的学习算法就是误差反向传播(error back propagation,BP)算法.这种算法基于最速下降法原理,主要缺点是学习速度过慢.超限学习机(extreme learning machine,ELM)极大地优化了单隐层神经网络的学习速度,却需要更多的隐层单元来达到与BP网络相当的效率,这不可避免地使网络结构冗余、测试时间变长.受到一种结合了ELM和最速下降法思想的USA(upper-layer-solution-aware)算法的启发,提出一种基于共轭梯度法的单隐层神经网络快速算法,并把它应用于不同数据库中.试验结果表明,在相同网络结构情况下,本算法的效率要优于ELM和USA算法.展开更多
文摘Accurate handwriting recognition has been a challenging computer vision problem,because static feature analysis of the text pictures is often inade-quate to account for high variance in handwriting styles across people and poor image quality of the handwritten text.Recently,by introducing machine learning,especially convolutional neural networks(CNNs),the recognition accuracy of various handwriting patterns is steadily improved.In this paper,a deep CNN model is developed to further improve the recognition rate of the MNIST hand-written digit dataset with a fast-converging rate in training.The proposed model comes with a multi-layer deep arrange structure,including 3 convolution and acti-vation layers for feature extraction and 2 fully connected layers(i.e.,dense layers)for classification.The model’s hyperparameters,such as the batch sizes,kernel sizes,batch normalization,activation function,and learning rate are optimized to enhance the recognition performance.The average classification accuracy of the proposed methodology is found to reach 99.82%on the training dataset and 99.40%on the testing dataset,making it a nearly error-free system for MNIST recognition.
文摘In this paper, classical and continuous variable (CV) quantum neural network hybrid multi-classifiers are presented using the MNIST dataset. Currently available classifiers can classify only up to two classes. The proposed architecture allows networks to classify classes up to n<sup>m</sup> classes, where n represents cutoff dimension and m the number of qumodes on photonic quantum computers. The combination of cutoff dimension and probability measurement method in the CV model allows a quantum circuit to produce output vectors of size n<sup>m</sup>. They are then interpreted as one-hot encoded labels, padded with n<sup>m</sup> - 10 zeros. The total of seven different classifiers is built using 2, 3, …, 6, and 8-qumodes on photonic quantum computing simulators, based on the binary classifier architecture proposed in “Continuous variable quantum neural networks” [1]. They are composed of a classical feed-forward neural network, a quantum data encoding circuit, and a CV quantum neural network circuit. On a truncated MNIST dataset of 600 samples, a 4-qumode hybrid classifier achieves 100% training accuracy.
文摘单隐层前馈神经网络是一种高效且结构简单的神经网络,它的一种典型的学习算法就是误差反向传播(error back propagation,BP)算法.这种算法基于最速下降法原理,主要缺点是学习速度过慢.超限学习机(extreme learning machine,ELM)极大地优化了单隐层神经网络的学习速度,却需要更多的隐层单元来达到与BP网络相当的效率,这不可避免地使网络结构冗余、测试时间变长.受到一种结合了ELM和最速下降法思想的USA(upper-layer-solution-aware)算法的启发,提出一种基于共轭梯度法的单隐层神经网络快速算法,并把它应用于不同数据库中.试验结果表明,在相同网络结构情况下,本算法的效率要优于ELM和USA算法.
