Currently, most studies on content caching strategies for vehicular networks rely the Zipf distribution model of popularity. This approach, however, often fails to accommodate time-varying nature popularity and traffic. Although some have begun considering popularity, they overlook implications vehicle traffic, which significantly influences allocation policies could potentially reduce edge server rental costs service providers. Addressing this gap, our study takes both traffic data into...

Feedforward-designed convolutional neural network (FF-CNN) is an interpretable network. The parameter training of the model does not require backpropagation (BP) and optimization algorithms (SGD). entire based on statistical data output by previous layer, parameters current layer are obtained through one-pass manner. Since complexity under FF design lower than BP algorithm, FF-CNN has better utility method in directions semi-supervised learning, ensemble continuous subspace learning....

Graph neural networks based on message-passing mechanisms have achieved advanced results in graph classification tasks. However, their generalization performance degrades when noisy labels are present the training data. Most existing labeling approaches focus visual domain or node tasks and analyze impact of only from a utility perspective. Unlike work, this paper, we measure effects noise data privacy model perspectives. We find that degrade model's enhance ability membership inference...

As one of popular machine learning algorithms, decision trees can be applied to both classification and regression tasks. Many tree models are trained used by some companies researchers on data obtained from users for business decision. However, the use may pose a potential risk privacy when training contains sensitive personal information. In this paper, we proposed an effective differentially private algorithm based Pearson's correlation coefficient, called Diff-PCCDT. The exponential...

Deep learning has been extensively applied in many fields, such as image segmentation, voice recognition, automatic language translation. However, malicious attackers attempt to attack the model which was trained accomplish a deep assignment via various schemes. Recently, differential privacy technology proposed defend against attacks sacrificing accuracy of model. Therefore, optimization methods have reduce overall cost, and aim seek tradeoff between utility. In this paper, we propose an...

A novel signal processing method based on calibration is present in this paper to improve the resolution of quadrature encoder. radial magnetizing ring alnico adhered rotor motor, while six Hall ICs are placed around evenly. magnetic filed produced by can be sensed. Quadrature encoder calibrated previously get table using a higher optical off-line. designed and its achieve 13-bits. Structure very simple cost low compare with an It used unfavorable environment where high-accuracy position...

In recent years, graph neural networks (GNNs) have become a popular semi-supervised learning method for processing graph-structured data. However, traditional network models rely heavily on labeled data during while neglecting the potential of unlabeled To further exploit data, pseudo-labeling-based methods select high-confidence pseudo-labeling training and assign nodes to join accordingly. Unfortunately, there is confidence bias in process with this approach. On other hand, self-supervised...

The feedforward-designed convolutional neural network (FF-CNN) method was recently proposed by Kuo et al. It has strong interpretability and low training complexity. In this paper, we have two improvements (1) We merge algorithms Layer-wise Relevance Propagation (LRP) FF-CNN to build an interpretable framework called LFB-CNN. back-propagation (BP) algorithm is used train the fully connected layer of FF-CNN. Meanwhile, LRP decompose calculate correlation between input output layer, further...

Successive subspace learning, a novel green, unsupervised and interpretable machine learning paradigm is widely used in image point cloud data classification, medical diagnosis, forgery detection. The multi-stage Saab transform Channel-wise extract low-level to high-level features for successive but the feature extraction process has some privacy security issues that hinder its further development. Li et al. [1] have proposed corresponding privacy-preserving schemes transform. However, above...