With the rapid development of autonomous driving technology, traffic sign and road surface classification have become key tasks for enhancing driving safety and comfort. Progress in this area is closely tied to the development of Convolutional Neural Network (CNN) for visual scene understanding. However, despite convolutional neural networks achieving human-level accuracy, their high computational and storage demands present significant challenges for real-time deployment on embedded devices. To address this, we propose the Elastic Reparameterization Neural Architecture Search (ERNAS) method, designed to efficiently adapt across multiple hardware platforms. ERNAS consists of two core modules: elastic reparameterization super-network design and hardware-aware multi-objective search, which collectively optimize both model performance and resource efficiency. Extensive experiments show that this method not only achieves high accuracy across various hardware platforms but also meets real-time performance requirements. Notably, we successfully implemented real-time inference for traffic sign recognition and road surface classification on the K210 platform, which has only 1.5MB of on-chip cache, achieving both high accuracy and an average real-time frame rate of approximately 35 FPS.

The abuse of community detection algorithms may bring the risk of privacy leakage. To protect personal privacy in complex networks, community detection attack algorithms are proposed, which can hide the true community structure of the whole network from the community detection algorithms by adding and deleting subtle edges. However, most of the existing algorithms perform attack based on a community structure so that a specific community detection method is usually adopted for obtaining the communities, which causes the algorithms to not perform well when the attacked community detection algorithm is unknown. To this end, a local structure-based community detection attack heuristic approach (LSHA) is proposed in this article, where the local structures, including several nodes with dense connections instead of the whole community structures, are considered. Unlike the whole community structures obtained by different community detection algorithms, which are usually different, the nodes in such a local structure are often assigned into the same community so that the attack is more general for different community detection algorithms. Specifically, in LSHA, a local structure selection strategy is proposed to maximize the attack effect, which selects two local structures for rewiring attack. Furthermore, two metrics, i.e., edge vulnerability and node entropy, are also suggested to select the nodes and edges for attack. In the experiments, the proposed LSHA is compared with five state-of-the-art attack algorithms. The experimental results against five representative community detection algorithms on nine real-world networks show that the proposed algorithm LSHA achieves good performance on both the attack effectiveness and the efficiency.

Identifying functional modules in protein-protein interaction (PPI) networks elucidates cellular organization and mechanism. Various methods have been proposed to identify the functional modules in PPI networks, but most of these methods do not consider the noisy links in PPI networks. They achieve a competitive performance on the PPI networks without noisy links, but the performance of these methods considerably deteriorates in the noisy PPI networks. Furthermore, the noisy links are inevitable in the PPI networks. In this paper, we propose a novel link-driven label propagation algorithm (LLPA) to identify functional modules in PPI networks. The LLPA first find link clusters in PPI networks, and then the functional modules are identified from the link clusters. Two strategies aimed to ensure the robustness of LLPA are proposed. One strategy involves the proposed LLPA updating the link labels in accordance with the designed weight of the link, which can reduce the incidence of noisy links. The other strategy involves the filtration of some noisy labels from the link clusters to further reduce the influence of noisy links. The performance evaluation on three real PPI networks shows that LLPA outperforms other eight state-of-the-art detection algorithms in terms of accuracy and robustness.

Wavelet and reflectivity inversion (WRI) is an important issue in seismic data processing. To overcome the ill-posedness of WRI inversion with more efficient parameter selection and better lateral continuity of reflectivities, we propose a new WRI algorithm named compact smoothness and relative sparsity (CSRS) algorithm, where a normalized compact constraint and a normalized smooth regularization is proposed for the wavelet inversion, and a relative sparsity constraint is proposed for the reflectivity inversion. The proposed constraints and regularization make the parameters of WRI easy to be selected. The proposed relative sparsity constraint can lead to a reflectivity profile with good lateral continuity as it can be suitable for various seismic data with a fixed sparsity parameter. We also propose an efficient algorithm for solving corresponding WRI optimization problem. The whole WRI problem is divided into reflectivity inversion subproblem and wavelet inversion subproblem by using alternating iterative method, where the initial wavelet is estimated by smoothing the absolute amplitude spectrum of averaged seismic data. The proximal algorithm is applied to solve both reflectivity inversion subproblem and wavelet inversion subproblem. By replacing Toeplitz matrix multiplication with fast Fourier transform (FFT) and using compact wavelet, our algorithm can be efficient for 3D seismic data. The numerical examples on 2D synthetic data, 2D offshore field data and 3D onshore field data demonstrate that, compared to Toeplitz-sparse matrix factorization (TSMF) algorithm, the CSRS algorithm with fixed default parameters can get high-resolution reflectivities with better lateral continuity, and requires much less computational time.

As a frequently-used dimensionality reduction technique in machine learning, feature selection has attracted interests in the last decade. Since feature selection is essentially a combinatorial optimization problem, how to search the valuable feature subset is a challenging optimization task. Particle swarm optimization (PSO) algorithm and its variations have shown their competitiveness in solving feature selection problem. However, they have been proven to be easily trapped into the local optimal in high-dimensional space due to their intrinsic characteristic of quick convergence. To this end, an effective binary particle swarm optimization algorithm, named CBPSOFS, is proposed for feature selection, where a community-grouping based adaptive updating strategy is designed to avoid trapping into the local optimum and enhance the performance of PSO algorithm in feature selection. To be specific, the correlationship among features is used to construct the feature network, where multiple feature groups are obtained by dividing the achieved feature network. Considering that a community usually contains multiple similar features, the proposed adaptive updating strategy utilizes these feature groups to make the similar features not be included in the same particle so as to maintain the diversity of the population in the evolution. In addition, an information gain based initialization strategy and a history information based resetting strategy are also developed to improve the quality of obtained feature subset. Experimental results on several real world datasets have demonstrated the effectiveness of CBPSOFS in feature selection when compared with the several state-of-the-art baselines.

Earthquakes are one of the most costly natural disasters facing human beings, which happens without an explicit warning, therefore earthquake prediction becomes a very important and challenging task for humanity. Although many existing methods attempt to address this task, most of them use either seismic indicators (explicit features) designed by geologists, or feature vectors (implicit features) extracted by deep learning methods, to characterize an earthquake for earthquake prediction. The problem of combining these two kind of features to improve final earthquake prediction performance remains pretty much open. To this end, we propose a deep learning model named DLEP to effectively fuse the explicit and implicit features for accurate earthquake prediction. In DLEP, we adopt eight precursory pattern-based indicators as the explicit features, and use a convolutional neural network (CNN) to extract implicit features. Then, an attention-based strategy is suggested to fuse these two kinds of features well. In addition, a dynamic loss function is designed to deal with the category imbalance of seismic data. Finally, experimental results on eight datasets from different regions demonstrate the effectiveness of the proposed DLEP for earthquake prediction comparing to several state-of-the-art baselines.