Multi-label text classification attempts to assign a label set to one specific document, which is more closely related to real life. Network models based on traditional deep learning achieve good prediction results. However, these models generally ignore the importance of label semantics and do not fit the connection between categories and text features well. Therefore, this paper proposes a novel L-RCap model. L-RCap uses the Bi-LSTM to extract global text features. With the global text features, we can use the label semantics to construct label-text features in the label semantic attention mechanism. Besides, we use the capsule network to extend the features information and use the dynamic routing algorithm to fit the association between features and categories. Compared with the baseline models, our model exhibits the best performance on two datasets.

With the development of the times, people generate a huge amount of data every day, most of which are unlabeled data, but manual labeling needs a lot of time and effort, so unsupervised algorithms are being used more often. This paper proposes an unsupervised image clustering algorithm based on contrastive learning and K-nearest neighbors (CLKNN). CLKNN is trained in two steps, which are the representation learning step and the clustering step. Contrastive learning and K-nearest neighbors have a huge impact on CLKNN. In the representation learning step, firstly CLKNN processes the image by double data augmentation to get two different augmented images; then CLKNN uses double contrastive loss to extract the high-level feature information of the augmented images, maximizing the similarity of row space and maximizing the similarity of column space to ensure the invariance of information. In the clustering step, CLKNN finds the nearest neighbors of each image by K-nearest neighbors, then it maximizes the similarity between each image and its nearest neighbors to get the final result. To test the performance of CLKNN, the experiments are conducted on CIFAR-10, CIFAR-100 and STL-10 in this paper. From the final results, it is clear that CLKNN has better performance than other advanced algorithms.

In recent years, a new attempt has been made to capture label co-occurrence by applying the sequence-to-sequence (Seq2Seq) model to multi-label text classification (MLTC). However, existing approaches frequently ignore the semantic information contained in the labels themselves. Besides, the Seq2Seq model is susceptible to the negative impact of label sequence order. Furthermore, it has been demonstrated that the traditional attention mechanism underperforms in MLTC. Therefore, we propose a novel Seq2Seq model with a different label semantic attention mechanism (S2S-LSAM), which generates fused information containing label and text information through the interaction of label semantics and text features in the label semantic attention mechanism. With the fused information, our model can select the text features that are most relevant to the labels more effectively. A combination of the cross-entropy loss function and the policy gradient-based loss function is employed to reduce the label sequence order effect. The experiments show that our model outperforms the baseline models.

Reinforcement learning (RL) has been successfully applied to solve path planning problems, but learning is generally slow. The main reason is not making full use of information collected during interaction with the environment. This paper proposes a novel method to solve the discrete space path planning problem in an environment without prior knowledge with intensive obstacles based on RL and heuristic search. Firstly, we apply Dyna-Q algorithm of RL to explore the map and search for the target point and optimize its policy with upper confidence bound (UCB). Then, when the target point is found, we use heuristic search to plan the path from the starting point to the target point and narrow the path to a small range. Finally, we combine Dyna-Q algorithm with the heuristic search recommended path for path planning. We evaluate our algorithm using maze navigation problem. The results verify that heuristic search accelerates Dyan-Q convergence.

The classic inverted residual block has been used in the design of existing lightweight neural network models, but it still has a large number of parameters and calculations, and there is room for improvement in the accuracy and speed of the models. In this paper, we propose a novel bottleneck structure design, called the Efficient Mobile Inverted Residual Block. Based on our block, a new lightweight mobile computer vision architecture is presented, called MobileNeSt, that has small parameters and high accuracy. MobileNeSt improves state-of-the-art performance of mobile models and is suitable for classification and detection tasks. Lightweight convolutions and new bottleneck structure design are used to compress and accelerate the model effectively, shortcut connection and a novel attention mechanism are added to improve the accuracy of the model. The proposed model has been extensively benchmarked on multiple public datasets. On CIFAR100 dataset, the Top-1 accuracy of MobileNeSt is improved by 1.1% over MobileNetV2, while the parameters of the model are reduced by 46.8%, the calculations are reduced by 30.4%, and the speed is increased by 7%. Compared with other computational-efficient architectures, our MobileNeSt achieves better results with fewer parameters.

Abstract. In recent years, a number of detectors have been proposed to improve the accuracy and speed of object detection tasks. However, poor detection performances for small objects and difficulties in optimizing deep networks remain critical challenges for object detection. We try to tackle these problems in two ways. First, we propose an innovative cross-fusion block (CFB) module that can enhance the representational power of features for instances of small objects. In CFBs, high-level features with rich semantic information and low-level features from different layers at the same scale are cross-fused together. Second, we propose a periodic oscillation attenuation learning rate (POA_lr) that can effectively skip some purely locally optimal solutions in the training process to obtain better detection accuracy. Extensive experiments on PASCAL VOC and MS COCO datasets show that CFB and POA_lr can achieve higher detection accuracy while maintaining real-time processing speeds. The code will be made publicly available.

With the increasing demand for application scenarios such as autonomous driving and drone aerial photography, it has become a challenging problem that how to achieve the best trade-off between segmentation accuracy and inference speed while reducing the number of parameters. In this paper, a lightweight and efficient asymmetric network (LEANet) for real-time semantic segmentation is proposed to address this problem. Specifically, LEANet adopts an asymmetric encoder-decoder architecture. In the encoder, a depth-wise asymmetric bottleneck module with separation and shuffling operations (SS-DAB module) is proposed to jointly extract local and context information. In the decoder, a pyramid pooling module based on channel-wise attention (CA-PP module) is proposed to aggregate multi-scale context information and guide feature selection. Without any pre-training and post-processing, LEANet respectively achieves the accuracy of 71.9% and 67.5% mean Intersection over Union (mIoU) with the speed of 77.3 and 98.6 Frames Per Second (FPS) on the Cityscapes and CamVid test sets. These experimental results show that LEANet achieves an optimal trade-off between segmentation accuracy and inference speed with only 0.74 million parameters.

YOLOv4-tiny is a simplified version of YOLOv4 detector, which is extremely fast and with few parameters. However, the detection performance of YOLOv4-tiny is poor while the recognition of small targets and occluded objects is weak. It is mainly attributed to the lack of feature extraction ability and learning ability of the backbone. Furthermore, the feature pyramid network (FPN) cannot adequately fuse adjacent feature maps in the process of multi-scale feature fusion. In this paper, a real-time detector with a bi-channel feature fusion mechanism is proposed based on YOLOv4-tiny, called BFF-YOLO, which effectively improves the detection of small targets and occluded objects. BFF-YOLO is composed of two main components: feature extraction and feature fusion. In the process of feature extraction, inspired by the idea of cross-stage partial connections (CSP), an enhanced CSP block (ECSPBlock) is proposed for enhancing the feature extraction of the backbone and the learning capability of the network. Moreover, the Maxpool layer in YOLOv4-tiny, which is used for downsampling and tends to lose fine-grained information, is replaced with a convolutional layer. In the process of feature fusion, a bi-channel feature fusion pyramid network (BFPN) is proposed to adequately fuse adjacent feature maps of different scales so that each detection head has both shallow and deep features. Finally, with a small increase in parameters, BFF-YOLO has achieved 36.5% AP and 85.1% mAP on the COCO and VOC datasets, respectively.

引入了Inception-SE卷积模块组来提升LeNet-5网络的广度与深度,运用SE模块增强了有用的特征并抑制了对当前任务用处不大的特征;使用BN层和Dropout优化网络,防止梯度弥散,提升精度;使用全局池化层(global average pooling,GAP)代替全连接层来减少网络计算参数.研究结果表明:改进后网络的识别精度达到了99.88％,比传统的LeNet-5网络提高了1.71％.

针对传统板形模式识别方法存在精度低、鲁棒性弱的问题,提出了一种混合优化RBF-BP组合神经网络板形模式识别方法.首先利用自组织映射网络(SOM)对样本聚类,利用聚类后的网络拓扑结构确定RBF的中心,并计算RBF的宽度,克服了传统聚类算法随机选取中心导致聚类结果不稳定的问题.然后利用遗传算法(GA)良好的全局搜索能力优化整个网络的权值.RBF-BP组合神经网络是由一个RBF子网和一BP子网串联构成的,该网络同时具备BP神经网络能较好地预测未知样本的能力以及RBF神经网络的逼近速度快的优点.并以某900HC可逆冷轧机板形识别为应用背景,在MATLAB2010a环境下进行仿真实验,结果表明混合优化RBF-BP组合神经网络的板形模式识别方法能够识别出常见的板形缺陷,提高了板形缺陷识别精度并具有较好的鲁棒性,可以满足板带轧机高精度的板形控制要求.

Typical convolutional filter only extract features linearly. Although nonlinearities are introduced into the feature extraction layer by using activation functions and pooling operations, they can only provide point-wise nonlinearity. In this paper, a Gaussian convolution for extracting nonlinear features is proposed, and a hybrid nonlinear convolution filter consisting of baseline convolution, Gaussian convolution and other nonlinear convolutions is designed. It can efficiently achieve the fusion of linear features and nonlinear features while preserving the advantages of traditional linear convolution filter in feature extraction. Extensive experiments on the benchmark datasets MNIST, CIFAR10, and CIFAR100 show that the hybrid nonlinear convolutional neural network has faster convergence and higher image recognition accuracy than the traditional baseline convolutional neural network.

常规单RBF神经网络板形识别模型不能全面分离出输入变化对每个特征参数的影响,为此设计了多RBF神经网络板形识别模型,用多个子网络分别识别不同的特征参数,能够更直接、更充分的提取出输入与每个输出的关系.仿真研究结果表明:所设计的多RBF神经网络板形识别模型能够正确识别出全部板形缺陷的类型,并且识别精度上比单RBF神经网络板形识别模型提高了16.1%.

针对常规神经网络板形识别方法中存在的不足,提出了以PCA替代欧氏距离作为提取特征的手段,并将所设计的PCA?RBF板形识别模型以FPGA为硬件实现载体进行了仿真研究.仿真结果表明,设计的PCA?RBF板形识别模型能够正确识别出板形缺陷,网络结构比常规RBF板形识别模型相对简化,同时识别精度提升了59％,抗干扰能力提升了82％.FPGA仿真结果在精度和实时性上可以满足实际工程需要.

Aiming at the complex feature extraction process of traditional handwritten Chinese character recognition,low recognition rate and weak discrimination ability of general deep learning classification model,a multi-channel cross-fusion deep residual network model and improves the central loss function designed in this paper. Firstly,based on the original data set pre-processing to reduce the risk of the model in over-fitting;then,the improved central loss function and the Softmax loss function are combined as the monitoring signal of model training,which can effectively aggregate data sets within the class,disperse between classes and improve the classifica-tion performance of the model in the training process;finally,the pre-processed data sets are input into the model designed in this pa-per. The experiment on CASIAHWDB-V1. 1 data set shows that the recognition model and algorithm designed in this paper can effec-tively improve the recognition rate of handwritten Chinese characters.

在公路交通中,针对复杂环境下交通标志识别率不高的问题,提出了一种基于K-means对图像聚类,切割图像感兴趣区域(Regions of Interest,ROI),并利用方向梯度直方图特征(Histogram of Oriented Gradient,HOG)与卷积运算,特征加权(CNN-Squeeze)相结合的交通标志识别方法.首先,采用K-means对交通标志图像进行三角形、圆形图像二聚类,并利用制作的切割模板切割ROI并提取HOG特征;然后,利用卷积神经网络(Convolutional Neural Network,CNN)对HOG特征进行过滤、降维,并通过Squeeze网络对过滤后的二次特征进行重要性标定;最后,训练该网络模型并实现对交通标志的识别.仿真结果表明,与BP网络、SVM及CNN对比,本文方法在保证训练时间的同时,识别精度达到98.58％.

Object detection is a challenging computer vision problem with numerous practical applications. Due to low accuracy and slow detection speed in object detection, we propose a real-time object detection algorithm based on YOLOv3. First, to solve the problem that features are likely to be lost in the feature extraction process of YOLOv3, a DB-Darknet-53 feature extraction network embedded in inception structure is designed, which effectively reduces the loss of features. Second, the detection network of YOLOv3 and the reuse of deep features in multiscale detection network are improved. Finally, the numbers and sizes of anchor boxes are selected by K-means clustering analysis, and the detection model is obtained by means of multiscale training. The improved algorithm has a mean average precision of 0.835 on the PASCAL VOC data set and a detection speed of 35.8 f/s, which is better than YOLOv3. (C) 2019 SPIE and IS&T

In view of the complicated background of the fire image, the complicated process of extracting the artificial feature, the poor generalization ability of the fire image, the low accuracy, false alarm rate, missing rate, the novel method for detecting fire images of multilinear principal component analysis (MPCA) was presented in the paper. The fire image recognition model was established by using MPCANet. Through the MPCA algorithm, the learning filter was used as the convolution kernel of deep learning network convolution layer, and the feature extraction of high dimensional images of tensor objects was taken, and candle images and fireworks images were taken as interference. Compared with other fire image recognition methods, the recognition accuracy of the proposed image recognition method reaches 97.5%, false alarm rate of 1.5%, missing rate of 1%. Experiments results show that this method could effectively solve the problems of fire image recognition.

针对传统优化算法(SNPOM)在辨识RBF-ARX模型参数时易陷入局部最优解的问题,将云遗传算法(CGA)和SNPOM算法结合,提出一种混合优化算法CGA-SNPOM.并以某公司900HC可逆冷轧机板形识别为应用背景,设计了基于CGA-SNPOM优化RBF-ARX的板形缺陷识别模型.分别用SNPOM算法和CGA-SNPOM算法对RBF-ARX模型参数进行优化,仿真验证表明,基于CGA-SNPOM优化的板形识别系统克服了SNPOM容易陷入局部极值的缺点,识别精度大幅提高,是一种有效的板形识别方案.

针对传统火灾报警系统存在着准确度不高、误报、漏报及泛化能力不强的问题,设计了一种基于GA(Genetic Algorithm)优化T-S云推理网络火灾探测模型,对模型进行训练和测试.并将T-S云推理网络与模糊神经网络对火灾信号的识别结果进行对比,给出MATLAB的仿真结果.通过仿真结果表明,该火灾探测模型识别精度更高,偏离样本数据测试结果与期望输出之间误差更小,模型提高了对火灾信号的识别精度和泛化能力.