Search

Previous convolutional neural network (CNN) based blind super-resolution (SR) methods usually adopt an iterative optimization way to approximate the ground-truth (GT) step-by-step. This solution always involves more computational costs to bring about time-consuming inference. At present, most blind SR algorithms are dedicated to obtaining high-fidelity results; their loss function generally employs L1 loss. To further improve the visual quality of SR results, perceptual metric, such as NIQE, is necessary to guide the network optimization. However, due to the non-differentiable property of NIQE, it cannot be as the loss function. Towards these issues, we propose an adaptive modulation network (AMNet) for multiple degradations SR, which is composed of the pivotal adaptive modulation layer (AMLayer). It is an efficient yet lightweight fusion layer between blur kernel and image features. Equipped with the blur kernel predictor, we naturally upgrade the AMNet to the blind SR model. Instead of considering iterative strategy, we make the blur kernel predictor trainable in the whole blind SR model, in which AMNet is well-trained. Also, we fit deep reinforcement learning into the blind SR model (AMNet-RL) to tackle the non-differentiable optimization problem. Specifically, the blur kernel predictor will be the actor to estimate the blur kernel from the input low-resolution (LR) image. The reward is designed by the pre-defined differentiable or non-differentiable metric. Extensive experiments show that our model can outperform state-of-the-art methods in both fidelity and perceptual metrics.

The prosody is an integration of the formal beauty and the content beauty of Chinese classical poetry. In this study, a prosodic structure prediction method based on a pretrained language representation model was proposed by us. Based on the pretrained language representation model, a separate output layer was set for each prosody level, with character as the modeling unit. Then the model was fine-tuned with the labeled prosody data. To achieve the simultaneous prediction of different prosodic levels in the input text, a word segmentation task was additionally introduced and the multitask learning method was used to model the relationship between the multilevel prosody and lexicon words. The experimental results verify that adding the word segmentation task can further improve model performance and demonstrate that the proposed method considerably reduces the demand for training data while maintaining an excellent prediction performance.

Wind power is a kind of green energy, renewable, pollution-free, inexhaustible, so wind power has a broad market prospects, its technology is also in continuous progress, wind turbine assembly unit capacity is higher and higher, the scale of wind farm is also towards large-scale development. As the capacity of wind turbine assembly increases, the blade diameter of the matching wind turbine needs to be increased correspondingly to obtain more wind energy. Taking offshore wind turbine as a representative, the blade diameter of some wind turbines has reached more than 100 meters. When the blade diameter becomes larger, it will bring a series of problems. The main performance is that the elasticity and flexibility of wind turbine tower, transmission shaft, blades and other main components will increase correspondingly, while the damping value of power electronic components will be smaller. These problems make the fatigue load variable of large wind turbine more complex, which will lead to the low frequency mode of the unit The more complex the modes are, the more complex they are. When the diameter of the wind turbine increases, the pitch angle increases due to tower shadow effect, wind shear and turbulence time-varying factors, which causes the blade to swing and shimmy more significantly, and then causes the fatigue load condition of the wind turbine to be more obvious, which seriously affects the mechanical stress intensity change between various transmission mechanisms of the wind turbine, thus reducing the service life of the wind turbine Limit. With the increase of blade diameter, the frequency of grid flicker will also be caused, which will cause voltage instability and reduce the power quality of grid after the wind turbine is connected to the grid. With the increase of impeller diameter and fan capacity, in recent years, independent variable pitch control technology came into being.Independent pitch control technology not only has the characteristics of centralized pitch control, but also can balance the unbalanced load on variable speed wind turbine, so as to reduce the fatigue load of wind turbine. Therefore, it has been paid attention to in the field of wind power research in the world. There are many research results in this area, but the research on the relationship between independent pitch control strategy and grid flicker is relatively less. The factors causing voltage fluctuation include air density, impeller speed, pitch angle, wind field distribution and power grid condition. This paper analyzes the wind shear and tower shadow effect which cause the vibration of wind turbine, studies the relationship between independent pitch strategy and wind turbine voltage change, proposes a control strategy of independent pitch and designs a controller to control the power of wind turbine during operation The method of voltage variation to make it stable, and then reduce the flicker of power grid, and carry out simulation experiments to verify its effectiveness. The control strategy can reduce the grid flicker at the common point of wind turbine outlet and grid connection, improve the stability of wind turbine grid connection, and improve the overall power quality of the grid.

With the progress and development of socio-economic technology, the contradiction between the strong demand for energy in various industries and the increasing shortage of traditional petrochemicals is deepening. People are increasingly aware of the importance of developing renewable energy, especially the use and development of wind energy. Pitch control is the core technology of wind power generation. At present, most of the mainstream products in the international wind power market apply centralized variable pitch control. The advantage of centralized pitch control is that it has a high wind energy utilization coefficient, which can ensure the maximum power in the low wind speed section, the rated power in the high wind speed section, improve the starting performance and braking performance of the wind turbine, improve the overall flexibility of the wind turbine, Force conditions of small machines and blades. Independent pitch control technology is a control technology method developed in recent years with the continuous increase of fan capacity and the constant increase of impeller diameter.This paper takes the wind energy conversion system of large doubly-fed induction generators as the research object, proposes an independent pitch control strategy, uses FAST and MATLAB / Simulink for dynamic modeling, simulation and comprehensive analysis. The results of simulation experiments show that The unbalanced load of wind turbine units is an effective control method to improve the power supply quality and extend the service life of wind turbine units.

This paper reviewed the 3rd NTIRE challenge on single-image super-resolution (restoration of rich details in a low-resolution image) with a focus on proposed solutions and results. The challenge had 1 track, which was aimed at the real-world single image super-resolution problem with an unknown scaling factor. Participants were mapping low-resolution images captured by a DSLR camera with a shorter focal length to their high-resolution images captured at a longer focal length. With this challenge, we introduced a novel real-world super-resolution dataset (RealSR). The track had 403 registered participants, and 36 teams competed in the final testing phase. They gauge the state-of-the-art in real-world single image super-resolution.

Deep convolutional neural networks (DCNN) have recently advanced the state-of-the-art on the issue of single image super-resolution (SR). In this work, we propose a two-stage convolutional network (TSCN) to estimate the desired high-resolution (HR) image from the corresponding low-resolution (LR) image. Specifically, we propose the multi-path information fusion (MIF) module that collects abundant information from feature maps of the input, output and intermediary in a module and distills primary information therein. Several cascaded MIF modules are used to progressively extract features desired by reconstruction and the output of each module is gathered for rebuilding the HR image. In addition, we introduce a refinement network with local residual topology architecture as the second stage so as to further restore the high-frequency details of HR image produced by the first stage. Due to less number of filters, the compact model achieves fast inference time and brings about state-of-the-art SR results on four benchmark datasets simultaneously. Code is available at https://github.com/Zheng222/TSCN.

In recent years, datasets represented in multi-view contain more information because different views describe different aspects. In fact, the structure of the dataset is embedded in a union of certain low-dimensional subspaces. Therefore, multiview subspace learning is a powerful technology to find the underlying structure and cluster data points correctly. Actually, the information contained in different views is different. Furthermore, the original data may be noisy. However, most existing multi-view subspace clustering methods treat each view equally and learn the self-expressiveness coefficient matrix of each view on the original data, which would decrease clustering performance. To solve the above problems, we propose a new method which learns a latent representation in an adaptive way. Meanwhile, the local geometrical structure is maintained on the latent representation by graph regularization. At the same time, a basic subspace clustering method is performed on the latent representation to get self-expressiveness coefficient matrix. We formulate the above problems into a unified optimization framework. Experimental results on several real-world datasets show the effectiveness of the proposed method.

Recently, deep convolutional neural networks (CNNs) have been demonstrated remarkable progress on single image super-resolution. However, as the depth and width of the networks increase, CNN-based super-resolution methods have been faced with the challenges of computational complexity and memory consumption in practice. In order to solve the above questions, we propose a deep but compact convolutional network to directly reconstruct the high resolution image from the original low resolution image. In general, the proposed model consists of three parts, which are feature extraction block, stacked information distillation blocks and reconstruction block respectively. By combining an enhancement unit with a compression unit into a distillation block, the local long and short-path features can be effectively extracted. Specifically, the proposed enhancement unit mixes together two different types of features and the compression unit distills more useful information for the sequential blocks. In addition, the proposed network has the advantage of fast execution due to the comparatively few numbers of filters per layer and the use of group convolution. Experimental results demonstrate that the proposed method is superior to the state-of-the-art methods, especially in terms of time performance., Comment: To appear in CVPR2018

The finite element model of sliding wear is established by applying finite element theory according to Archard wear model. The general program is developed by ANSYS Parametric Design Language (APDL). The wear analysis of self-lubricating joint bearing liner is carried out. The simulation and the experimental results are very consistent with the relative error being 6.38%. In the simulation, the concept of the wear step length is introduced, and the method of moving boundary key point of the liner region in the model is used to describe the material removal process. The parametric modeling is used to refine the liner layer and then divide the mesh, and then the internal meshes distortion caused by boundary point change is solved. Meanwhile, the appropriate wear step length is used to improve the computational efficiency and accuracy. The results show that the finite element method can predict the sliding wear from the viewpoint, which provides a practical tool for practical engineering wear design.

Principal Component Analysis (PCA), one of the most popular dimensionality reduction algorithms, has three particular problems: the number of eigenvalues is limited by the two direction dimensions; it assumes for reconstruction of Gaussian distributed data, not for classification problems; it assumes that eigenvalues and eigenvectors is all linear. In this paper, we proposed a Bayesian Mixture Model, Bayesian Mixture of Inverse Regression (BMI), to deal with these three problems as preprocessing method and then use classic algorithms, Discriminative Locality Alignment (DLA) and Fishers Linear Discriminant Analysis (FLDA), to classify the test data into different topics. Through empirical studies on the face recognition demonstrate the effectiveness of DLA & BMI and LDA & BMI are more effective than DLA & PCA and LDA & PCA.

The network dynamic evaluation is in the proven prescription legal science has truthfully given the trusted degree of the network from the real network to the simulation experiment evaluation concrete model and the realization process. The network modelling process uses the extraction real network data, establishes the simulation environment, under the simulation environment destructive collapse test method in the dynamic test process. Useing the result data to carries on the sampling test in the real network, to confirmation model establishment accuracy and the evaluation result credibility.

The application of MODIS produces Inner Mongolia snow-cover map is a blank research area in China. The validation result shows that the snow-cover map was accurate. MODIS data monitors large-scale snow cover is very effective.

In order to evaluate the ability of personnel training of the university accurately, combined talents teaching and research condition in engineering colleges and universities, this paper constructs a comprehensive evaluation index system of cultivation of innovative talents in engineering colleges and universities and propose a method of comprehensive evaluation based on Particle Swarm Optimization and support vector machine through analysis on the strength, activity and effectiveness of cultivation of innovative talents of university. 46 evaluation index system of the 26 universities is carried out classification test, evaluation results show that the model has good regression results, and also suitable for evaluation grade of cultivation of innovative talents in engineering colleges and universities at present.

In this paper, we apply the least-square support vector machine (LS-SVM) to operon prediction of Escherichia coli (E.coli), with different combinations of intergenic distance, gene expression data, and phylogenetic profile. Experimental results demonstrate that the WO pairs tend to have shorter intergenic distances, higher correlation coefficient and much stronger relation of co-envoled between phylogenetic profiles. Also, we dealt with the data sets extracted from WOs? and TUBs?, processed the intergenic distances with log-energy entropy, de-noised the Pearson correlation coefficients of two genes expression data with wavelet transform, and computed the Hamming distances of two phylogenetic profiles. Then we trained LS-SVM using part of the data sets and tested the trained classifier model using the rest data sets. It shows that different combinations of features could affect the prediction results. When the combination of intergenic distance, gene expression data and phylogenetic profile is taken as the input of LS-SVM in the linear kernel type, good results can be obtained, of which the accuracy, sensitivity and specificity are 92.34%, 93.54%, and 90.73%, respectively.

In order to evaluate efficiency in university laboratory in a reasonable way, the index system for efficiency evaluation is established, and the efficiency class is separated into three classes-good, fair, and poor. To classify the efficiency of three classes, the evaluation model of multi-layer support vector machines (SVM) classifier is established. In order to verify the effectiveness of the method, 41 universities laboratories which are chosen from 211 project are used, and Levenberg-Marquardt neural network is also used to classify the same data and make comparison. The experiment results show that multi-layer SVM classifier is effective in efficiency evaluation for university laboratory when the training data set is not too much, and the method achieves better performance than Levenberg - Marquardt neural network.