Epilepsy is a brain disorder caused by abnormal discharges of neurons in brain. It is one of the most commonly studied disorders in neurology. The research of epilepsy electroencephalogram (EEG) has become a hot resea...
详细信息
With the increasing performance of deep convolutional neural networks, they have been widely used in many computer vision tasks. However, a huge convolutional neural network model requires a lot of memory and computin...
详细信息
ISBN:
(数字)9798350349184
ISBN:
(纸本)9798350349191
With the increasing performance of deep convolutional neural networks, they have been widely used in many computer vision tasks. However, a huge convolutional neural network model requires a lot of memory and computing resources, which makes it difficult to meet the requirements of low latency and reliability of edge computing when the model is deployed locally on resource-limited devices in edge environments. Quantization is a kind of model compression technology, which can effectively reduce model size, calculation cost and inference delay, but the quantization noise will cause the accuracy of the quantization model to decrease. Aiming at the problem of precision loss caused by model quantization, this paper proposes a post-training quantization method based on scale optimization. By reducing the influence of redundant parameters in the model on the quantization parameters in the process of model quantization, the scale factor optimization is realized to reduce the quantization error and thus improve the accuracy of the quantized model, reduce the inference delay and improve the reliability of edge applications. The experimental results show that under different quantization strategies and different quantization bit widths, the proposed method can improve the accuracy of the quantized model, and the absolute accuracy of the optimal quantization model is improved by 1.36%. The improvement effect is obvious, which is conducive to the application of deep neural network in edge environment.
The issue of maximizing the influence is a hot topic in the research of social network. Many researchers have studied from the perspective of the structure of the network such as the LeaderRank algorithm. However, the...
详细信息
ISBN:
(纸本)9781728111421;9781728111414
The issue of maximizing the influence is a hot topic in the research of social network. Many researchers have studied from the perspective of the structure of the network such as the LeaderRank algorithm. However, the algorithm lacks semantic interpretability and explanations for user behavior. Therefore, we propose a novel URI (user-relational iterative) rank to address the above issues. The URI rank is divided into two parts to obtain the values of user influence. The first part is the forwarding probability based on the user relationship. We introduce the relationship between users to the user transition probability and use the random forest to quantify the value of the forwarding probability. The second part is the random transition probability based on the ground node. We optimize the weight assigning of the random transition probability by combining static decentralization and dynamic decentralization. Thus, we more accurately represent the user's random transfer behavior. The experiments performed on the Sina Micro-Blog Dataset show that our algorithm outperforms the existing algorithms.
Aiming at the problem that the result of some attack sequence alignment methods is not necessarily the optimal expression of their characteristics. This paper presents a Production Rule Sequence Alignment Algorithm (P...
详细信息
Aiming at the problem that the result of some attack sequence alignment methods is not necessarily the optimal expression of their characteristics. This paper presents a Production Rule Sequence Alignment Algorithm (PRSA) combining the production rule inference mechanism which improves traditional sequence alignment algorithm. A new accumulation of knowledge is obtained by changing the way of sequence alignment and the transformation of signatures. PRSA overcomes the problem that the extraction results produced by the traditional sequence alignment algorithm cannot express the attack signature accurately. Then, we establish an automatic attack signature generation model based on PRSA. The experimental results show that the matching results obtained by using PRSA can express the signatures of the attack accurately and improve the detection rate of the attacks.
In a Loss of Coolant Accident (LOCA), reactor core temperatures can rise rapidly, leading to potential fuel damage and radioactive material release. This research presents a groundbreaking method that combines the pow...
详细信息
Autonomous driving systems require real-time environmental perception to ensure user safety and experience. Streaming perception is a task of reporting the current state of the world, which is used to evaluate the del...
详细信息
The pre-training language model BERT has brought significant performance improvements to a series of natural language processing tasks, but due to the large scale of the model, it is difficult to be applied in many pr...
The pre-training language model BERT has brought significant performance improvements to a series of natural language processing tasks, but due to the large scale of the model, it is difficult to be applied in many practical application scenarios. With the continuous development of edge computing, deploying the models on resource-constrained edge devices has become a trend. Considering the distributed edge environment, how to take into account issues such as data distribution differences, labeling costs, and privacy while the model is shrinking is a critical task. The paper proposes a new BERT distillation method with source-free unsupervised domain adaptation. By combining source-free unsupervised domain adaptation and knowledge distillation for optimization and improvement, the performance of the BERT model is improved in the case of cross-domain data. Compared with other methods, our method can improve the average prediction accuracy by up to around 4% through the experimental evaluation of the cross-domain sentiment analysis task.
Defect detection aims to locate the accurate position of defects in images, which is of great significance to quality inspection in the industrial product manufacturing. Currently, many defect detection methods rely o...
详细信息
Defect detection aims to locate the accurate position of defects in images, which is of great significance to quality inspection in the industrial product manufacturing. Currently, many defect detection methods rely on deep neural networks to extract features. Although the accuracy of these methods is relatively high, it is computationally intensive, making the methods difficult to deploy in resource-limited edge devices. In order to solve these problems, a lightweight defect detection model for the industrial edge environment is proposed, termed the efficient defect detection network (EDDNet). EfficientNet-B0 is used as the feature extraction backbone, extracting feature maps from feature layers of different depths of the network and fusing multilevel features by multilevel feature fusion (MFF). To obtain more information, we redesign the attention mechanism in MBConv blocks, taking the encoding space (ES) attention mechanism as a new module, which solves the problem that the defective image spatial information is ignored. The experimental results on the NEU-DET and DAGM2007 datasets and PCB defect datasets demonstrate the effectiveness of the proposed EDDNet and its possibility for application in industrial edge device.
Large language models (LLMs), with their powerful generative capabilities and vast knowledge, empower various tasks in everyday life. However, these abilities are primarily concentrated in high-resource languages, lea...
详细信息
Thanks to the emergence of transformers and Vision Transformer (VIT), attention mechanisms have also been applied to medical image registration. However, the current attention mechanisms in medical image registration ...
详细信息
暂无评论