Search

Neural communication is a promising communication technique in future nano-scale applications. In neural communication systems, neurons are able to change their synaptic strength over time, which is known as synaptic plasticity. The plasticity may affect the synaptic strength, and hence have significant influence on information transmission efficiency in neural communications. In this paper, we focus on a neural communication system under synaptic plasticity including vesicle release and post-synaptic spike generation processes. Specifically, one typical learning rule of the synaptic plasticity, i.e., spike-time dependent plasticity (STDP), and the related consolidation effect are considered, where the synaptic weights can be enhanced or decreased depending on temporal correlations between presynaptic spike arrival and postsynaptic firing. The mutual information for the neural communication system is theoretically analyzed considering the influence of STDP, which is also evaluated through simulations under various factors. It is observed that STDP can increase the mutual information between synapses under realistic models of vesicle release probabilities and the synaptic noise. When the initial synaptic strength is relatively large, the consolidation effect helps improve the communication rates under both single-input single-output (SISO) and multi-input single-output (MISO) channels.

Detecting road surface defects is essential for highway maintenance, yet the application of most models is hindered by the limitations of existing detection resources. To address this challenge, we have enhanced YOLOv8, introducing a lightweight detection model dubbed QL-YOLOv8s. In this study, we employ the DIoU loss function to optimize bounding box regression, taking into account both the size of overlapping areas and the distance between the centers of boxes, thereby handling targets of various sizes and shapes with improved localization accuracy. Moreover, a lightweight Mixed Local Channel Attention (MLCA) has been incorporated into the backbone of the model, aimed at enhancing the recognition capabilities in complex environments without in-creasing the model’s burden. Furthermore, by integrating the Dilated Wrapping Residual (DWR) module and C2f into BiFPN, we developed a new neck structure, BiFPN-D, and introduced a lightweight detection head, Detect-T3, thus augmenting the model’s feature perception capacity, reducing parameter count, and boosting detection speed. Based on the RDD 2022 public dataset, QL-YOLOv8s demonstrated a reduction in parameter count and size by 37%, a decrease in com-putational requirements by 19%, and achieved an average precision of mAP0.5 at 95.8%. These results underscore the contribution and practical value of our method to the technology of automatic road defect detection.

The detection of pavement diseases is an important and basic link in the road maintenance process. Many methods based on deep learning have been applied. However, these methods are not accurate enough and cannot accurately identify defects in complex background with shadow occlusion and uneven lighting brightness. In order to overcome the shortcomings of previous detection methods, a complex background defect detection algorithm based on improved YOLOv7 is proposed. First, the K-means++ clustering algorithm is used for initial anchor box setting to obtain better anchor box parameters; then, the group spatial pyramid pooling module SPPCSPC_G is introduced to replace the original SPPCSPC module to improve the fusion speed of image features and thereby improve the detection accuracy; Finally, the GELU activation function is used as the activation function of the REPConv convolution module in the YOLOv7 model, which effectively reduces model overfitting and thereby improves model detection accuracy. The test results show that the average accuracy of the improved detection algorithm for disease detection increased from 65.4% to 72.3%, an increase of 6.9%, the amount of calculation and parameters decreased by 4% and 14.9% respectively, and the FPS reached 80, an increase of 17%, and no pavement defects are missed or wrongly detected. It is more suitable for real-time detection of defects in complex background. It can be seen that the improved YOLOv7 has better detection effect on complex background defects.

In the process of production, automobile steel forgings are prone to various cracks, which affect the product quality. At present, forgings defects are mainly detected by fluorescent magnetic particle inspection and manual inspection. Aiming at the problems of low detection accuracy and efficiency in this method, an improved convolutional neural network model is proposed. The fluorescent magnetic particle inspection images of two typical forgings were intelligently inspected. Firstly, a deep learning model with EfficientNet as the backbone and Feature Pyramid Network (FPN) as the fusion layer is constructed. Secondly, in order to improve the convergence speed and detection accuracy, the calculation method of intersection over union is improved, and the network is improved by using the Attention Mechanism. Finally, Particle Swarm Optimization algorithm (PSO) with adaptive parameters is introduced to optimize the hyperparameters of neural network, and a fluorescent magnetic particle inspection image acquisition platform is built for verification. The mean Average Precision (mAP) of the best model of EfficientNet-PSO on the validation set is 95.69%. F1 score is 0.94 and FLOPs is 1.86B. Compared with other five deep learning neural network models, this method effectively improves the defect detection efficiency and accuracy of flange plate and cylinder head, which can meet the defect detection requirements.

Properly wearing a face mask has become an effective way to limit the COVID-19 transmission. In this work, we target at detecting the fine-grained wearing state of face mask: face without mask, face with wrong mask, face with correct mask. This task has two main challenging points: 1) absence of practical datasets, and 2) small intra-class distance and large inter-class distance. For the first challenging point, we introduce a new practical dataset covering various conditions, which contains 8635 faces with different wearing status. For the second challenging point, we propose a novel detection framework about conditions of wearing face mask, named Context-Attention R-CNN, which enlarge the intra-class distance and shorten inter-class distance by extracting distinguishing features. Specifically, we first extract the multiple context feature for region proposals, and use attention module to weight these context feature from channel and spatial levels. And then, we decoupling the classification and localization branches to extract more appropriate feature for these two tasks respectively. Experiments show that the Context-Attention R-CNN achieves 84.1% mAP on our proposed dataset, outperforming Faster R-CNN by 6.8 points. Moreover, Context-Attention R-CNN still exceed some state-of-the-art single-stage detectors.

With the development of recurrent neural networks (RNN), various natural language generation (NLG) tasks have boomed in the past few years, such as response generation in conversation and poetry generation. However, automatic generation of news comments is anew, challenging and not well-studied task in NLG. Different from other NLG tasks, this task requires the contextual relevance between comments and news. In addition, we need to generate diversified comments, because different people usually have different opinions on the same news in the real world. In this paper, we propose a gated attention neural network model (GANN) to generate news comments. To address the problem of contextual relevance, we introduce the gated attention mechanism to use news context self-adaptively and selectively. To ensure the diversity of comments, we use random sample and relevance control to generate comments with different topics and degrees of relevance. Moreover, we apply generative adversarial nets to improve GANN. Automatic evaluation with perplexity score reveals that GANN outperforms the existing comment generation methods. Human evaluation proves that the generated news comments are close to human comments.

For the control request of specialty vehicles, this paper combines a phase plane analysis with the variable universe fuzzy control technique. Moreover, we develop a new systemic design strategy for the adaptive fuzzy logic controller. With the phase plane analysis, the complete rule base that consists of few key rules can be objectively established, which avoids the irrationality introduced into the process via the subjectivity of the designer of the fuzzy controllers. Furthermore, with the variable universe, the design requirements of the membership functions can be relaxed. Meanwhile, the accuracy of the performance of the fuzzy logic controller can be enhanced despite the limited number of fuzzy rules in the rule base. Based on the Lyapunov stability theory, the stability of the close-loop system can be guaranteed. Simulation results of a double inverted pendulum demonstrate the feasibility of the design strategy for the adaptive fuzzy logic controller, which simplifies the design of the fuzzy logic controller and ensures control. The application of this design strategy can significantly lighten the burden for fuzzy logic controller designers and shorten the development period of the fuzzy logic controller.