Search

Abstract Tuning the oxygen activity in perovskite oxides (ABO3) is promising to surmount the trade-off between activity and selectivity in redox reactions. However, this remains challenging due to the limited understanding in its activation mechanism. Herein, we propose the discovery that generating subsurface A-site cation (Lasub.) vacancy beneath surface Fe-O layer greatly improved the oxygen activity in LaFeO3, rendering enhanced methane conversion that is 2.9-fold higher than stoichiometric LaFeO3 while maintaining high syngas selectivity of 98% in anaerobic oxidation. Experimental and theoretical studies reveal that absence of Lasub.-O interaction lowered the electron density over oxygen and improved the oxygen mobility, which reduced the barrier for C-H bond cleavage and promoted the oxidation of C-atom, substantially boosting methane-to-syngas conversion. This discovery highlights the importance of A-site cations in modulating electronic state of oxygen, which is fundamentally different from the traditional scheme that mainly credits the redox activity to B-site cations and can pave a new avenue for designing prospective redox catalysts.

Abstract Convolutional neural networks (CNN) have a strong feature extraction ability for images and present a high level of efficiency and accuracy in object detection and image recognition. When CNN is used to model microwave devices, the existing literature generally uses its size parameters as one‐dimensional (1‐D) input, which does not give full play to the image‐processing ability of CNN. In order to make full use of the characteristics of CNN, this letter converts the 1‐D input of microwave devices into the form of an image model, that is, the 1‐D input is transformed into a two‐dimensional (2‐D) matrix composed of 0 and 1 as the input. The image model is combined with CNN, called image‐based CNN (ICNN), which establishes a deep learning surrogate model between the physical parameters and electrical properties of microwave devices and improves the accuracy and generalization ability of the model. Taking the resonant frequency of the microstrip antenna as a simulation example, modelling was carried out by the proposed ICNN and compared with the mainstream machine learning methods. The results show that the proposed method has high convergence and fitting accuracy.

Abstract Because of extreme sensitivity to time and energy consumption, many computation‐ and data‐intensive tasks are difficult to implement on mobile terminals and cannot meet the needs of the rapid development of mobile networks. To solve this problem, mobile edge computing (MEC) appears to be a promising solution. In this study, we propose two offloading schemes in the multiple unmanned aerial vehicles (UAVs) enabled MEC network. Their optimisation goals are to minimise the global computing time and energy consumption of all UAVs, respectively. Different from previous research, the UAV can perform tasks locally or offload an appropriate percentage to the desired MEC server in the two proposed schemes. In order to get the minimum global computing time, we prove the existence condition and obtain the optimal offloading proportion. In addition, in order to minimise global energy consumption, we also obtain the optimal offloading proportion and present the optimal transmission power through solving Karush–Kuhn–Tucker conditions. Finally, because UAVs are selfish, we adopt the game theory to get optimal solutions of the proposed offloading strategies. Numerical results verify that the proposed schemes can effectively decrease the global computing time and energy consumption, especially for a large number of UAVs.

When modeling and optimizing electromagnetic components, it is the most time consuming for obtaining the training samples with labels from full-wave electromagnetic simulation software. The traditional machine learning (ML) model is usually effective in the training process, unfortunately, its generalization ability is limited for practical application. Inspired by the artificial neural network (ANN) and the Gaussian process (GP) kernel learning model, a deep kernel learning (DKL) model with multiple-nonlinear-mapping layers is proposed based on particle swarm optimization (PSO) algorithm. In this work, correlation characteristics of training samples are transformed by multiple layers, and then they act as inputs of the GP model. Simultaneously, we use PSO to optimize the weights, biases of each mapping layer, and GP hyperparameters, aiming to determine and optimize the DKL model structure. As a result, the proposed DKL is suitable for processing small samples, as well as has the generalization ability of a deep network, which avoids the phenomenon of gradient disappearance to some extent caused by error back propagation (BP) in the deep network and improves the modeling accuracy while effectively ensuring the modeling efficiency. In this study, the performance of the DKL model is evaluated by using the resonant frequencies of two microstrip antennas (MSAs), and the predicted results of the DKL model are compared with those via different modeling methods. The results show that prediction accuracy of the DKL is 18.906% higher than that of the GP, 38.926% higher than the ANN, and 46.660% higher than the neural network ensemble (NNE).

In order to overcome the extremely time-consuming drawback of deep learning (DL), broad learning system (BLS) was proposed as an alternative method. This model is simple, fast, and easy to update. To ensure the fitting and generalization ability of BLS, the hidden layer neurons are often set too many, in fact, a lot of neurons are not needed. Greedy BLS (GBLS) is proposed in this paper to deal with the redundancy of the hidden layer in BLS from another perspective. Different from BLS, the structure of GBLS can be seen as a combination of unsupervised multi-layer feature representation and supervised classification or regression. It trains with a greedy learning scheme, performs principal component analysis (PCA) on the previous hidden layer to form a set of compressed nodes, which are transformed into enhancement nodes and then activated by nonlinear functions. The new hidden layer is composed of all newly generated compressed nodes and enhancement nodes, and so on. The last hidden layer of the network contains the higher-order and abstract essential features of the original data, which is connected to the output layer. Each time a new layer is added to the model, and there is no need to retrain from the beginning, only the previous layer is trained. Experimental results demonstrate that the proposed GBLS model outperforms BLS both in classification and regression.

For antenna modeling and optimization design, high-fidelity full-wave electromagnetic simulation software can generally be used to obtain training samples. However, this process takes a long time. To solve this problem, a two-stage Gaussian process (GP) considering electromagnetic sensitivity information is proposed. Based on the coarse grid and fine grid of electromagnetic simulation software, the first stage learns the mapping relationships of antenna performance and sensitivity of each parameter, simultaneously. In the second stage, the accurate surrogates are established based on the mapping relationships above obtained. Because the method in this study takes sensitivity information into consideration during the modeling process, it can better reflect the mapping relationship between antenna input and output, which can develop a more accurate surrogate model. The proposed two-stage GP surrogate model considering sensitivity information can significantly reduce the demand of high-fidelity training samples for modeling, and greatly save the simulation time of calling electromagnetic simulation software. Therefore, this method is more suitable for the problems of insufficient electromagnetic simulation samples. The proposed approach is evaluated by modeling and optimization of an inverted F antenna and an ultra-wideband planar monopole antenna. The results show that the method has high modeling accuracy with limited training samples given by high-fidelity electromagnetic simulation, which further verify its effectiveness and efficiency.

Due to the absorption and scattering effects of light propagating through water, underwater images inevitably suffer from severe degradation, such as color casts and losses of detail. Many existing deep learning-based methods have demonstrated superior performance for underwater image enhancement (UIE). However, accurate color correction and detail restoration still present considerable challenges for UIE. In this work, we develop a dual-branch fusion network, dubbed the DBFNet, to eliminate the degradation of underwater images. We first design a triple-color channel separation learning branch (TCSLB), which balances the color distribution of underwater images by learning the independent features of the different channels of the RGB color space. Subsequently, we develop a wavelet domain learning branch (WDLB) and design a discrete wavelet transform-based attention residual dense module to fully employ the wavelet domain information of the image to restore clear details. Finally, a dual attention-based selective fusion module (DASFM) is designed for the adaptive fusion of latent features of the two branches, in which both pleasing colors and diverse details are integrated. Extensive quantitative and qualitative evaluations of synthetic and real-world underwater datasets demonstrate that the proposed DBFNet significantly improves the visual quality and shows superior performance to the compared methods. Furthermore, the ablation experiments demonstrate the effectiveness of each component of the DBFNet.

Gaussian process (GP) is a rapidly developing supervised machine learning (ML) method in recent years, which has been widely used in the establishment of surrogate models in the field of electromagnetics. However, it has the problems of large sample demand, high computational complexity and low accuracy when processing high dimensional data. To solve this problem, a manifold Gaussian process (MGP) ML method based on differential evolution (DE) algorithm is proposed in this study. For the proposed method, the DE algorithm is used to get dimension reduction parameters, and the method can work very well with the optimized parameters. Compared with the traditional GP model, the dimensionality reduction method based on Isomap is adopted to simplify the mapping relationship between data pairs. Therefore, the model is more suitable for the problem of insufficient samples and high data dimension. In this study, the proposed DE-based MGP (DE-MGP) is applied to the extraction of coupling coefficients of the fourth-order and sixth-order coupling filters, in which the test error of the fourth-order coupling filter surrogate model can be reduced to 0.84%, and the test error of the sixth-order coupling filter is expected to be reduced to 1.53%, which proves that the proposed method is very effective.

For the optimal design of electromagnetic components, surrogate model methods can usually be used, but obtaining labeled training samples from full-wave electromagnetic simulation software is most time-consuming. How to use relatively few labeled samples to obtain a relatively high-precision surrogate model is the current electromagnetic research hotspot. This article proposes a semi-supervised co-training algorithm based on Gaussian process (GP) and support vector machine (SVM). By using a small number of initial training samples, the initial GP model and initial SVM model can be trained by some basic parameter settings. Moreover, the accuracy of these two models can be improved by using the differences between these two models and combining with unlabeled samples for jointly training. In the co-training process, to ensure the performance of the proposed algorithm, a stop criterion set in advance to control the number of unlabeled samples introduced. Therefore, the accuracy of the model can be prevented from being reduced by introducing too much unlabeled samples, which can find the best solution in the limited time. The proposed co-training algorithm is evaluated by benchmark functions, optimal design of Yagi microstrip antenna (MSA) and GPS Beidou dual-mode MSA. The results show that the proposed algorithm fits the benchmark functions well. For the problem of resonant frequency modeling of the above two different MSAs, under the condition of using the same labeled samples, the predictive ability of the proposed algorithm is improved compared with the traditional supervised learning method. Moreover, for the groups of antenna sizes that meet the design requirements, the fitting effects of their return loss curve (S11) are well. The effectiveness of the proposed co-training algorithm has been well verified, which can be used to replace the time-consuming electromagnetic simulation software for prediction.

As an image processing method, underwater image enhancement (UIE) plays an important role in the field of underwater resource detection and engineering research. Currently, the convolutional neural network (CNN)- and Transformer-based methods are the mainstream methods for UIE. However, CNNs usually use pooling to expand the receptive field, which may lead to information loss that is not conducive to feature extraction and analysis. At the same time, edge blurring can easily occur in enhanced images obtained by the existing methods. To address this issue, this paper proposes a framework that combines CNN and Transformer, employs the wavelet transform and inverse wavelet transform for encoding and decoding, and progressively embeds the edge information on the raw image in the encoding process. Specifically, first, features of the raw image and its edge detection image are extracted step by step using the convolution module and the residual dense attention module, respectively, to obtain mixed feature maps of different resolutions. Next, the residual structure Swin Transformer group is used to extract global features. Then, the resulting feature map and the encoder’s hybrid feature map are used for high-resolution feature map reconstruction by the decoder. The experimental results show that the proposed method can achieve an excellent effect in edge information protection and visual reconstruction of images. In addition, the effectiveness of each component of the proposed model is verified by ablation experiments.

For the optimal design of electromagnetic devices, it is the most time consuming to obtain the training samples from full wave electromagnetic simulation software, including HFSS, CST, and IE3D. Traditional machine learning methods usually use only labeled samples or unlabeled samples, but in practical problems, labeled samples and unlabeled samples coexist, and the acquisition cost of labeled samples is relatively high. This paper proposes a semisupervised learning Gaussian Process (GP), which combines unlabeled samples to improve the accuracy of the GP model and reduce the number of labeled training samples required. The proposed GP model consists two parts: initial training and self-training. In the process of initial training, a small number of labeled samples obtained by full wave electromagnetic simulation are used for training the initial GP model. Afterwards, the trained GP model is copied to another GP model in the process of self-training, and then the two GP models will update after crosstraining with different unlabeled samples. Using the same test samples for testing and updating, a model with a smaller error will replace another. Repeat the self-training process until a predefined stopping criterion is met. Four different benchmark functions and resonant frequency modeling problems of three different microstrip antennas are used to evaluate the effectiveness of the GP model. The results show that the proposed GP model has a good fitting effectiveness on benchmark functions. For microstrip antennas resonant frequency modeling problems, in the case of using the same labeled samples, its predictive ability is better than that of the traditional supervised GP model.

In order to reduce the time of designing microstrip antenna, this paper proposes a self-renewing fitness estimation of particle swarm optimization algorithm (SFEPSO) to improve the design efficiency. Firstly, a fitness predictive model of the particles is constructed according to the evolution formula of particle swarm optimization (PSO). From the third generation of the algorithm, the fitness of particles is given by the prediction model instead of the full-wave electromagnetic simulation. Aiming to keep the right direction of evolution, the prediction model will be proofed every N generation during the optimization process. If the prediction model accuracy is lower than the threshold, it will be updated and then continue iterating until the particles satisfy the demands. Compared with the traditional optimization method, this method greatly reduces the evaluation time and improves the design efficiency. The method is validated by the optimized design of E-type dual-frequency microstrip antenna and WLAN/WiMAX multiband antenna. The optimized results show that the purpose of rapid optimization can be achieved while ensuring the design accuracy.

We present a novel binary version of multiobjective wind driven optimization (WDO) for emitted beamforming of opportunistic array radar, which is assumed as a multiobjective optimization problem. Firstly, the emitted signal model and objective functions of optimization are presented. Then the algorithm proposes a new definition of the position vector of air parcel, and brings a good discretization interpretation of continuous WDO. For multiobjective optimization, the grey relational grade (GRG) is then used to measure the similarity between the best two solutions for these two objectives. The best pressure locations with the maximum GRG will be recorded as the best two candidate solutions to the problem, and a final optimization result will be selected according to the importance of the two objectives. Finally, the proposed improved WDO has been applied for the optimal design of beamforming of the opportunistic antenna array, which needs a trade-off between the 3 dB main beam width and sidelobe level. The simulation results show that the proposed method outperforms conventional particle swarm optimization (PSO) in the optimal beamforming by achieving more reduction in the sidelobe level and saving more runtime.

When modelling and optimizing antennas by machine learning (ML) methods, it is the most timeconsuming to obtain the training samples with labels from full-wave electromagnetic simulation software. To address the problem, this paper proposes an optimization method based on the consensus results of multiple independently trained Student's-T Process (STP) with excellent generalization ability. First, the STP is introduced as a surrogate model to replace the traditional Gaussian Process (GP), and the hyperparameters of the STP model are optimized. Afterwards, a consistency algorithm is used to process the results of multiple independently trained STPs to improve the reliability of the results. Furthermore, an aggregation algorithm is adopted to reduce the error obtained in the consistency results if it is greater than the consistency flag. The effectiveness of the proposed model is demonstrated through experiments with rectangular microstrip antennas (RMSA) and circular microstrip antennas (CMSA). The experimental results show that the use of multiple independently trained STPs can accelerate the antenna design optimization process, and improve modelling accuracy while maintaining modelling efficiency, which has high generalization ability. [ABSTRACT FROM AUTHOR]

The growth status of 352 malnourished and anemic children aged 0-3 years in Minhang District, Shanghai, were analyzed, and the effectiveness of health interventions was evaluated. The work focused on 352 malnourished and anemic children aged 0-3 years, summarized their health status by conducting two questionnaire surveys, and evaluated the effects of health interventions. The results revealed that the proportion of malnourished and anemic boys (57.10%) was slightly higher than that of girls (42.90%); and malnutrition-induced anemia accounted for 32.95% of the total number of children. The initial symptoms of malnourished children mainly manifested as below-normal height, weight loss, intellectual developmental delay and decreased muscle tone. Factors such as birth weight, number of infants, mode of delivery, preterm birth, feeding method, complementary feeding, medical history, and educational level of parents and primary caregivers had certain impacts on growth of children. Furthermore, the level of awareness of nutritional disorders knowledge was higher after the intervention compared to before the intervention. In conclusion, health education interventions for malnourished children can effectively improve their condition and enhance their overall health status, warranting widespread application and promotion. [ABSTRACT FROM AUTHOR]

Bayesian Optimization (BO) is an efficient global optimization algorithm, which is widely used in the field of engineering design. The probabilistic surrogate model and acquisition function are the two keys to the algorithm. Building an efficient probabilistic surrogate model and designing a collection function with excellent exploring capabilities can improve the performance of BO algorithm, allowing it to find the optimal value of the objective function with fewer iterations. Due to the characteristics of small samples and non-parametric derivation of the Gaussian Process (GP), traditional BO algorithms usually use the GP as a surrogate model. Compared with the GP, the Student's T Process (STP) retains the excellent properties of GP, and has more flexible posterior variance and stronger robustness. In this paper, STP is used as the surrogate model in BO algorithm, the hyperparameters of the model are optimized by STP, and the estimation strategy function (EST) is improved based on the posterior output of the optimized STP, thus realizing the improved BO algorithm based on the STP. To verify the performance of the proposed algorithm, numerical experiments are designed to compare the performances of the traditional BO algorithm, which includes the lower confidence bound function (LCB) and EST as acquisition function respectively and GP as the surrogate model, and the proposed BO algorithm with STP as the surrogate model and LCB, expected improvement function (EI), expected regret minimization function (ERM) as acquisition function respectively. The results show that the proposed algorithm in this paper performs well when finding the global minimum of multimodal functions. Based on the developed algorithm in this paper, the resonant frequency of printed dipole antenna and E-shaped antenna is modeled and optimized, which further confirms the good design ability and design accuracy of the BO algorithm proposed in this paper. [ABSTRACT FROM AUTHOR]