Entropy最新文献

Power-line presence recognition technology for unmanned aerial vehicles (UAVs) is one of the key research directions in the field of UAV remote sensing. With the rapid development of UAV technology, the application of UAVs in various fields has become increasingly widespread. However, when flying in urban and rural areas, UAVs often face the danger of obstacles such as power lines, posing challenges to flight safety and stability. To address this issue, this study proposes a novel method for presence recognition in UAVs for power lines using an improved GhostNet v2 knowledge distillation dual-attention mechanism convolutional neural network. The construction of a real-time UAV power-line presence recognition system involves three aspects: dataset acquisition, a novel network model, and real-time presence recognition. First, by cleaning, enhancing, and segmenting the power-line data collected by UAVs, a UAV power-line presence recognition image dataset is obtained. Second, through comparative experiments with multi-attention modules, the dual-attention mechanism is selected to construct the CNN, and the UAV real-time power-line presence recognition training is conducted using the SGD optimiser and Hard-Swish activation function. Finally, knowledge distillation is employed to transfer the knowledge from the dual-attention mechanism-based CNN to the nonlinear function and Mamba-enhanced GhostNet v2 network, thereby reducing the model's parameter count and achieving real-time recognition performance suitable for mobile device deployment. Experiments demonstrate that the UAV-based real-time power-line presence recognition method proposed in this paper achieves real-time recognition accuracy rates of over 91.4% across all regions, providing a technical foundation for advancing the development and progress of UAV-based real-time power-line presence recognition.

Simulating entangled, many-body quantum systems is notoriously hard, especially in the case of the high-dimensional nature of the underlying physical objects. In this work, we propose an approach for simulating the Potts model based on the Suzuki-Trotter decomposition that we construct for qudit systems. Specifically, we introduce two qudit-native decomposition schemes: (i) the first utilizes the Mølmer-Sørensen gate and additional local levels to encode the Potts interactions, while (ii) the second employs a light-shift gate that naturally fits qudit architectures. These decompositions enable a direct and efficient mapping of the Potts model dynamics into hardware-efficient qudit gate sequences for a trapped-ion platform. Furthermore, we demonstrate the use of a Suzuki-Trotter approximation with our evolution-into-gates framework for detecting the dynamical quantum phase transition. Our results establish a pathway toward qudit-based digital quantum simulation of many-body models and provide a new perspective on probing nonanalytic behavior in high-dimensional quantum many-body models.

Significant earthquakes can cause widespread infrastructure damage, social implications, and substantial economic losses. To mitigate these impacts, earthquake forecasting models have been developed to estimate earthquake occurrences and improve recovery efforts, with the Epidemic-Type Aftershock Sequence (ETAS) model being the most informative statistical framework for characterizing earthquake sequences. In this study, the ETAS model is used to estimate the model parameters for seismicity in Canada using the historical earthquake catalogue and to forecast long-term seismicity for seven different regions in Canada. Furthermore, the model is used to generate synthetic earthquake catalogues in order to assess its ability to replicate observed seismic patterns. The study identifies the southwestern region of Canada, associated with the coastal area of British Columbia, as being at the highest seismic risk, with a 66% exceedance probability for M7.5 events or above to occur in 30 years. In contrast, Alberta features the least seismic risk, with a 4% exceedance probability for events above 6.5 magnitude. For southeastern Canada, associated with Eastern Ontario and Southern Quebec, an exceedance probability of 74% for events above 6.0 magnitude poses the potential for significant damage due to the larger exposed population. Moreover, the resulting seismicity maps show the model's capability for real-events analysis, but improvements are needed for further applications.

We introduce Lehmer Activation Units (LAUs), a class of aggregation-based neural activations derived from the Lehmer transform that unify feature weighting and nonlinearity within a single differentiable operator. Unlike conventional pointwise activations, LAUs operate on collections of features and adapt their aggregation behavior through learnable parameters, yielding intrinsically interpretable representations. We develop both real-valued and complex-valued formulations, with the complex extension enabling phase-sensitive interactions and enhanced expressive capacity. We establish a universal approximation theorem for LAU-based networks, providing formal guarantees of expressive completeness. Empirically, we show that LAUs enable highly compact architectures to achieve strong predictive performance under tightly controlled experimental settings, demonstrating that expressive power can be concentrated within individual neurons rather than architectural depth. These results position LAUs as a principled, interpretable, and efficient alternative to conventional activation functions.

Accurate power load forecasting is crucial for ensuring the safety and economic operation of power systems. However, the complex, non-stationary, and heterogeneous nature of power load data presents significant challenges for traditional prediction methods, particularly in capturing instantaneous dynamics and effectively fusing multi-feature information. This paper proposes a novel framework-Ensemble Entropy with Adaptive Deep Fusion (EEADF)-for short-term multi-feature power load forecasting. The framework introduces an ensemble instantaneous entropy extraction module to compute and fuse multiple entropy types (approximate, sample, and permutation entropies) in real-time within sliding windows, creating a sensitive representation of system states. A task-adaptive hierarchical fusion mechanism is employed to balance computational efficiency and model expressivity. For time-series forecasting tasks with relatively structured patterns, feature concatenation fusion is used that directly combines LSTM sequence features with multimodal entropy features. For complex multimodal understanding tasks requiring nuanced cross-modal interactions, multi-head self-attention fusion is implemented that dynamically weights feature importance based on contextual relevance. A dual-branch deep learning model is constructed that processes both raw sequences (via LSTM) and extracted entropy features (via MLP) in parallel. Extensive experiments on a carefully designed simulated multimodal dataset demonstrate the framework's robustness in recognizing diverse dynamic patterns, achieving MSE of 0.0125, MAE of 0.0794, and R² of 0.9932. Validation on the real-world ETDataset for power load forecasting confirms that the proposed method significantly outperforms baseline models (LSTM, TCN, transformer, and informer) and traditional entropy methods across standard evaluation metrics (MSE, MAE, RMSE, MAPE, and R²). Ablation studies further verify the critical roles of both the entropy features and the fusion mechanism.

Accurate channel estimation is critical for enabling effective directional beamforming and spectrally efficient transmission in beamspace massive multiple-input multiple-output (MIMO) systems. However, conventional model-driven algorithms are derived from idealized mathematical models and typically suffer severe performance degradation under model mismatches caused by complex and nonideal propagation environments. Although data-driven deep learning (DL) approaches can learn channel characteristics from data, they typically require large-scale training datasets and demonstrate limited generalization capability. To overcome these limitations, we propose a model-data hybrid-driven network (MD-HDN) scheme to address the wideband beamspace channel estimation problem. In the MD-HDN scheme, we unfold the vector approximate message passing (VAMP) algorithm into a trainable network, where a novel shrinkage function is introduced to enhance the estimation accuracy. Extensive numerical results confirm that the proposed MD-HDN scheme can significantly outperform existing schemes under various signal-to-noise ratio (SNR), and achieve substantial improvements in both estimation accuracy and robustness.

The regional green patent cooperation network describes the structural characteristics of regional collaborative innovation, and the link prediction of the network can anticipate the overall evolution trend, as well as help organizations identify potential partners for technology collaboration. This paper proposes a link prediction model based on multidimensional features, which integrates prediction indicators of node features, path features, and content features. In the model, the entropy weight method is employed to integrate various node similarity indicators, the heterogeneous influence of intermediate links and nodes is incorporated to fully emphasize the issue of heterogeneous paths, and the content similarity feature indicator based on patent text topic analysis integrates multiple distance similarity metrics. To improve prediction accuracy, the Grey Wolf Optimizer (GWO) method is adopted to determine the optimal weights for the three-dimensional indicators. The comparative experimental results show that the multidimensional prediction model can improve prediction accuracy significantly. Finally, the proposed prediction model is applied to forecast the green patent cooperation network in the Beijing-Tianjin-Hebei region of China, and the prediction results are discussed based on the distribution of agent types and regional distribution.

Solving Bayesian inverse problems efficiently stands as a major bottleneck in scientific computing. Although Bayesian Physics-Informed Neural Networks (B-PINNs) have introduced a robust way to quantify uncertainty, the high-dimensional parameter spaces inherent in deep learning often lead to prohibitive sampling costs. Addressing this, our work introduces Quantum-Encodable Bayesian PINNs trained via Classical Ensemble Kalman Inversion (QEKI), a framework that pairs Quantum Neural Networks (QNNs) with Ensemble Kalman Inversion (EKI). The core advantage lies in the QNN's ability to act as a compact surrogate for PDE solutions, capturing complex physics with significantly fewer parameters than classical networks. By adopting the gradient-free EKI for training, we mitigate the barren plateau issue that plagues quantum optimization. Through several benchmarks on 1D and 2D nonlinear PDEs, we show that QEKI yields precise inversions and substantial parameter compression, even in the presence of noise. While large-scale applications are constrained by current quantum hardware, this research outlines a viable hybrid framework for including quantum features within Bayesian uncertainty quantification.

The identification of community structure is pivotal for understanding the functional characteristics of complex networks. To address the limitations of most existing community detection algorithms, which often require predefining the number of communities and lack robustness, this paper proposes a novel community detection algorithm named D-means (K-means community detection algorithm based on density peaks). This algorithm integrates the concept of density peak clustering with K-means spectral clustering, employing Chebyshev's inequality to automatically determine the number of community centers, thereby enabling unsupervised identification of community quantities. By designing a multi-dimensional evaluation framework, the comparative experiments were conducted on LFR benchmark networks (Lancichinetti-Fortunato-Radicchi benchmark networks) and real-world social network datasets. The results demonstrate that the D-means algorithm outperforms traditional algorithms in terms of ACC (accuracy), ARI (adjusted rand index), and NMI (normalized mutual information) metrics, while also achieving improvements in runtime efficiency, showcasing strong robustness. Finally, the D-means algorithm was applied to the public transportation network of Urumqi. Empirical analysis identified 12 functionally significant transportation communities, providing theoretical support for urban rail transit optimization and commercial facility layout planning.

The Profitable Tour Problem is a well-known NP-hard optimization challenge central to tourism planning, aiming to maximize collected profit while minimizing travel costs. While classical heuristics provide approximate solutions, they often struggle with finding globally optimal routes. This paper explores the application of near-term quantum computing to this problem. We propose a framework based on the Variational Quantum Eigensolver to find high-quality solutions for the Profitable Tour Problem. The core of our contribution is a novel methodology for constructing a constraint-aware variational ansatz that directly encodes the problem's hard constraints. This approach circumvents the need for large penalty terms in the Hamiltonian problem, which are often a source of optimization challenges. We validate our method through numerical simulations on a representative tourism scenario of up to 25 qubits. The results demonstrate the viability of the approach, achieving high solution accuracy consistent with brute-force enumeration for smaller instances. This work serves as a proof-of-concept for applying Variational Quantum Eigensolver to complex tourism optimization problems and provides a basis for future exploration on real quantum hardware.