Seismic travel time tomography using surface waves is an effective tool for three-dimensional crustal imaging. Historically, these surface waves are the result of active seismic sources or earthquakes. More recently, however, also surface waves retrieved through the application of seismic interferometry are exploited. Conventionally, two-step inversion algorithms are employed to solve the tomographic inverse problem. That is, a first inversion results in frequency-dependent, two-dimensional maps of phase velocity, which then serve as input for a series of independent, one-dimensional frequency-to-depth inversions. As such, a two-dimensional grid of localized depth-dependent velocity profiles are obtained. Stitching these separate profiles together subsequently yields a three-dimensional velocity model. Relatively recently, a one-step three-dimensional non-linear tomographic algorithm has been proposed. The algorithm is rooted in a Bayesian framework using Markov chains with reversible jumps, and is referred to as transdimensional tomography. Specifically, the three-dimensional velocity field is parameterized by means of a polyhedral Voronoi tessellation. In this study, we investigate the potential of this algorithm for the purpose of recovering the three-dimensional surface-wave-velocity structure from ambient noise recorded on and around the Reykjanes Peninsula, southwest Iceland. To that end, we design a number of synthetic tests that take into account the station configuration of the Reykjanes seismic network. We find that the algorithm is able to recover the 3D velocity structure at various scales in areas where station density is high. In addition, we find that the standard deviation on the recovered velocities is low in those regions. At the same time, the velocity structure is less well recovered in parts of the peninsula sampled by fewer stations. This implies that the algorithm successfully adapts model resolution to the density of rays. Also, it adapts model resolution to the amount of noise on the travel times. Because the algorithm is computationally demanding, we modify the algorithm such that computational costs are reduced while sufficiently preserving non-linearity. We conclude that the algorithm can now be applied adequately to travel times extracted from (time-averaged) station-station cross correlations by the Reykjanes seismic network.

For personalized recommender systems,matrix factorization and its variants have become mainstream in collaborative filtering.However,the dot product in matrix factorization does not satisfy the triangle inequality and therefore fails to capture fine-grained information. Metric learning-based models have been shown to be better at capturing fine-grained information than matrix factorization. Nevertheless,most of these models only focus on rating data and social information, which are not sufficient for dealing with the challenges of data sparsity. In this paper,we propose a metric learning-based social recommendation model called SRMC.SRMC exploits users' co-occurrence pattern to discover their potentially similar or dissimilar users with symmetric relationships and change their relative positions to achieve better recommendations.Experiments on three public datasets show that our model is more effective than the compared models.

Abstract: Image aesthetic assessment (IAA) with neural attention has made significant progress due to its effectiveness in object recognition. Current studies have shown that the features learned by convolutional neural networks (CNN) at different learning stages indicate meaningful information. The shallow feature contains the low-level information of images and the deep feature perceives the image semantics and themes. Inspired by this, we propose a visual enhancement network with feature fusion (FF-VEN). It consists of two sub-modules, the visual enhancement module (VE module) and the shallow and deep feature fusion module (SDFF module). The former uses an adaptive filter in the spatial domain to simulate human eyes according to the region of interest (ROI) extracted by neural feedback. The latter not only takes out the shallow feature and the deep feature by transverse connection, but also uses a feature fusion unit (FFU) to fuse the pooled features together with the aim of information contribution maximization. Experiments on standard AVA dataset and Photo.net dataset show the effectiveness of FF-VEN.

Although neural network architectures are critical for their performance, how the structural characteristics of a neural network affect its performance has still not been fully explored. We here map architectures of neural network to directed acyclic graphs, and find that incoherence, a structural characteristic to measure the order of directed acyclic graphs, is a good indicator for the performance of corresponding neural networks. Therefore we propose a deep isotropic neural network architecture by folding a chain of same blocks then connecting the blocks with skip connections at different distances. Our models, named FoldNet, have two distinguishing features compared with traditional residual neural netowrks. First, the distances between block pairs connected by skip connections increase from always equal to one to specially selected different values, which lead to more incoherent graphs and let the neural network explore larger receptive fields and thus enhance its multi-scale representation ability. Second, the number of direct paths increases from one to multiple, which leads to a larger proportion of shorter paths and thus improve the direct propagation of information throughout the entire network. Image classification results on CIFAR-10 and Tiny ImageNet benchmarks suggested that our new network architecture performs better than traditional residual neural networks.

Visual question answering (VQA) is receiving increasing attention from researchers in both the computer vision and natural language processing fields. There are two key components in the VQA task: feature extraction and multi-modal fusion. For feature extraction, we introduce a novel co-attention scheme by combining Sentence-guide Word Attention (SWA) and Question-guide Image Attention (QIA) in a unified framework. To be specific, the textual attention SWA relies on the semantics of the whole question sentence to calculate contributions of different question words for text representation. For the multi-modal fusion, we propose a “Cross-modal Multistep Fusion (CMF)” network to generate multistep features and achieve multiple interactions for two modalities, rather than focusing on modeling complex interactions between two modals like most current feature fusion methods. To avoid the linear increase of the computational cost, we share the parameters for each step in the CMF. Extensive experiments demonstrate that the proposed method can achieve competitive or better performance than the state-of-the-art.

In recent years, intelligent identification and prediction of ore deposits based on deep learning algorithm and image processing technology has gradually become one of the research frontiers in the field of geological and metallogenic prediction. However, this method also has many problems that need to be solved. For example, (1) very few trainable image samples containing mineral point labels; (2) The geological image features are small and irregular, and the image similarity is high; (3) It is difficult to calculate the influence of different geological prospecting factors on ore mineralization. Based on this, this paper constructs a deep learning network model Multiscale Feature Attention Framework (MFAF) based on geo-image data. The results show that MFCA-Net module in MFAF model can solve the problem of scarce mine label images to a certain extent. In addition, the channel attention mechanism SE-Net module can quantify the difference of influence of different source factors on mineralization. The prediction map is obtained by applying the MFAF model in the study of deposit identification and prediction in the research area of the southern section of Qin-hang metallogenic Belt. In the forecast map, many regions have high metallogenic potential. Through observation, it is found that the above predicted area covers 100% of the known ore deposits at present, which has a good prediction effect. Later, we can find out the specific reasons through field verification. The multi-scale feature fusion and attention mechanism MFA Framework in this paper can provide a new way of thinking for geologists in mineral exploration. The research of this paper also provides resource guarantee and technical support for the sustainable exploitation of mineral resources and the sustainable growth of society and economy.

In view of the problem that the hot and cold loads of mushroom houses are affected by multiple factors such as environment, mushrooms, and equipment, which are difficult to accurately model, this paper proposes a short-term load prediction method based on empirical wavelet transform (EWT) and mixed autoregressive integrated moving average (ARIMA) and convolutional bi-directional short- and long-term attention mechanism (CNN-BiLSTM-Attention). The method first uses the Boruta algorithm to filter the input characteristics, and then uses the EWT method to decompose the mushroom house load data into 4 modal components. Then the Lempel-Ziv method is introduced to divide the modal components into two categories of high and low frequencies, and the CNN-BiLSTM-Attention and ARIMA prediction models are constructed separately. Finally, the two types of prediction results are superimposed and reconstructed to obtain the final load prediction value. The test results show that the Boruta algorithm can effectively filter out the characteristics of key influencing factors. Compared with the Spearman and Pearson correlation coefficient methods, the mean absolute error (MAE) of the prediction results is reduced by 14.72% and 3.75%, respectively; compared with the ensemble empirical mode decomposition (EEMD) method, the EWT method can increase the decomposition and reconstruction error by 103 orders of magnitude, which can effectively improve the prediction accuracy of the model; compared with the individual neural network model, the prediction effect of the model proposed in this paper has obvious advantages, and the MAE, root mean square error (RMSE) and mean absolute percentage error (MAPE) of the prediction results are reduced respectively. 31.06%, 26.52% and 39.27%.

The aim of this study was to examine and understand the physiological and biochemical components of the blood and vitreous humor of goldfish (Carassius auratus). Blood was drawn from the arterial (static) line and vitreous humour was extracted using a syringe from healthy goldfish. An automatic haematology analyzer was used to detect 17 physiological indicators, including white blood cells and red blood cells. At the same time, 20 biochemical indexes, including albumin, calcium, and glutamic pyruvic transaminase, were found in the automatic biochemical analyzer. Experiments were also conducted on the effect of blood and vitreous fluid collection on the subsequent survival rate of the celestial goldfish. The findings demonstrated that there were values for 17 physiological indicators in the blood of the goldfish, including red blood cells (2.19 × 1012), white blood cells (62.21 × 109), hemoglobin (138.25g/L), and no eosinophils or basophils. However, no data for 17 physiological indications of vitreous humor were discovered. The quantities of total protein, glutamic oxaloacetic transaminase, and potassium in the vitreous humor were substantially lower than those in the serum (P < 0.01) according to the results of 20 biochemical indexes in the serum and vitreous humor of goldfish. Alkaline phosphatase, chlorine, and creatinine between vitreous humor and serum did not differ significantly(P > 0.05). The levels of glutamic pyruvic transaminase and γ-glutamyl transpeptide in vitreous humor were significantly higher than those in serum (P < 0.01). This experiment provides basic datas for the healthy culture of goldfish and the perfection of hematology in goldfish, and the results will be convenient for the further study of development and formation of celestial eye traits.

The unique characteristics of frescoes on overseas Chinese buildings can attest to the integration and historical background of Chinese and Western cultures. Reasonable analysis and preservation of overseas Chinese frescoes can provide sustainable development for culture and history. This research adopts the image analysis technology based on artificial intelligence, and proposes a ResNet-34 model and method integrating transfer learning. This deep learning model can identify and classify the source of the frescoes of the emigrants, and can effectively deal with the problems such as the small number of fresco images on the emigrants' buildings, poor quality, difficulty in feature extraction, and similar pattern text and style. The experimental results show that the training process of the model proposed in this article is stable. On the constructed Jiangmen and Haikou fresco JHD datasets, the final accuracy is 98.41%, and the recall rate is 98.53%. The above evaluation indicators are superior to classic models such as AlexNet, GoogLeNet, and VGGNet. It can be seen that the model in this article has strong generalization ability and is not prone to overfitting. It can effectively identify and classify the cultural connotations and regions of frescoes.

Understanding the thermal cracks of rock caused by high temperature is of great help to the development of underground engineering, such as geothermal mining, underground coal gasification. Although there has been many study on thermal cracking, it is not systematic and in-depth enough. In order to deeply study the thermal crack mechanism of rock, this paper conducted thermal cracking experiments on granite at different temperatures and heating/cooling paths, and qualitative and quantitative analysis was conducted on the crack initiation characteristics, propagation paths, and crack network morphology of rock thermal crack under different test conditions. The thermal crack mechanism was also analyzed from the perspectives of mineral petrology, fracture mechanics, thermodynamics, and other aspects. The research results show that there are two obvious mutation points in the study temperature range for samples with fast cooling paths (SF path: slow heating and fast cooling; FF path: fast heating and fast cooling), around 200-300 ℃ and 600 ℃, respectively, while for samples with slow cooling paths (SS path: slow heating and slow cooling; FS path: fast heating and slow cooling), there is only one mutation point around 600 ℃. The initiation positions of thermal cracking under all temperature paths are relatively similar, with intergranular crack located between particles at the edge of the sample or intercrystalline crack in the middle of feldspar or quartz aggregates. The initiation temperature of SF and FF path specimens is relatively low compared to SS and FS path specimens, and the number and size of cracks is small. The crack network structure formed by the SF path is the most complex, with the largest crack ratio and cumulative crack length; The crack network structure formed by the FF path is relatively complex, with a larger main crack size but relatively fewer secondary cracks; The FS and SS paths do not form a complex network with good connectivity. The process of thermal crack development can be divided into three stages: the development of small cracks, the joint development of main cracks and small cracks, and the connection of cracks into a network structure. The mechanism of thermal crack propagation is that the expansion and thermal conductivity between different crystals are different. The thermal stress caused by temperature gradient and the tension or shear stress caused by the inconsistent deformation of crystals form stress concentration in weak areas such as particles boundary, cleavage, and original cracks. Firstly, it causes some crystals with smaller strength or less rounded shape to crack, and when the combined stress is large, the cracks will gradually expand along the existing fine cracks. This also explains that the most main cracks of the SF and FF path specimens mainly surround some large mineral aggregates or between particles.

In view of the complexity of the pile foundation underpinning structure system and the stringent requirements of the construction process, this paper briefly describes the necessity of introducing epoxy resin reinforcing adhesive of planting rebar in the design of pile foundation underpinning beam structure to im-prove the mechanical properties of the underpinning beam new and old concrete joint surfaces and proposes a new type of pile foundation underpinning beam system construction method by “chiseling + prestressed reinforcement + epoxy resin reinforcing adhesive”. This paper uses an actual pile foundation underpinning project of an urban overpass as a prototype, designs and creates a model structure with a similarity ratio of 1/6, and performs repeated pro-gressive static loading tests to study the load carrying capacity, displacement change, and other properties of the underpinning structure, as well as analyses and distorts the overall working performance and failure mode of them. On this basis, the prototype structure's finite element analysis model was built, and the finite element analysis results were compared with the test results to get the mechanical properties and deformation characters of the actual pile foundation under-pinning structure system corresponding to the actual underpinning beam load. This paper's study can lay the theoretical and experimental foundation for the smooth development of similar projects.

A switchable and tunable terahertz (THz) metamaterial based on photosensitive silicon and Vanadium dioxide (VO2) was proposed. By using finite-difference time-domain (FDTD) method, the transmission and reflective properties of the metamaterial were investigated theoretically. The results imply that, the metamaterial can realize a dual electromagnetically induced transparency (EIT) or two narrow-band absorption depending on the temperature of the VO2. Additionally, the magnitude of the EIT and two narrow-band absorption can be tuned by varying the conductivity of photosensitive silicon (PSi) via pumping light. Correspondingly, the slow light effect accompanying the EIT can also be adjusted.

: This work introduces the instrument design of the medium energy proton detector (MEPD, detection range: 30keV-5MeV) mounted on the Chinese Fengyun-4B (FY-4B) satellite. Compared with the similar detector on the Fengyun-3E (FY-3E) satellite, the instrument has undergone significant changes due to the different orbital radiation environment and solar lighting conditions. Based on the calculation of radiation model AP8, the geometry factor is reduced to 0.002 cm2sr, while the MEPD on FY-3E satellite is 0.005 cm2sr. Another difference is that sensors in some directions is exposed to direct sunlight for 80 minutes every day on this orbit, depending on the attitude angle of the satellite, which is much worse than that on FY-3E satellite. According to the calculation results of permeability of photons through different materials, a 100 nm thickness nickel film is added in front of sensors in order to eliminate light pollution completely. The experiment on the solar simulator shows that the measure is effective and the detector has no error counts when the solar irradiance coefficient is 1.0. In addition, Geant4 software is applied to simulate particle transportation process under complete machine condition so that to check the contamination of electrons on sensors among all directions after magnetic deflection. Data obtained in orbit shows that the instrument works properly, and the data is in good agreement with the AP8 model. The observations of the MEPD on board the FY-4B satellite can provide important support for the safety of the spacecraft and the theoretical research related to space weather.

Ovarian cancer is the deadliest gynecological malignancy of the reproductive organs in the United States. Cyclin-dependent kinase 1 (CDK1) is an important cell cycle regulatory protein that spe-cifically controls the G2/M phase transition of the cell cycle. RO-3306 is a selective, ATP-competitive, and cell-permeable CDK1 inhibitor that shows potent anti-tumor activity in multiple pre-clinical models. In this study, we investigated the effect of CDK1 expression on the prognosis of patients with ovarian cancer and the anti-tumorigenic effect of RO-3306 in both ovarian cancer cell lines and a genetically engineered mouse model of high-grade serous ovarian cancer (KpB model). In 147 patients with epithelial ovarian cancer, overexpression of CDK1 was significantly associated with poor prognosis, compared with a low expression group. RO-3306 significantly inhibited cellular proliferation, induced apoptosis, caused cellular stress, and reduced cell migration. Treatment of KpB mice with RO-3306 for four weeks showed a significant decrease in tumor weight under obese and lean conditions without obvious side effects. Overall, our results demonstrate that inhibition of CDK1 activity by RO-3306 effectively reduces cell proliferation and tumor growth, providing biological evidence for future clinical trials of CDK1 inhibitors in ovarian cancer

Biogeography research is flawed by the poor understanding of microbial distributions due to the lack of a systematic research framework, especially regarding appropriate study units. By combining pure culture and molecular methods, we studied the biogeographic patterns of nematode-trapping fungi by collecting and analysing 2,250 specimens from 228 sites in Yunnan Province, China. We found typical watershed patterns at the species and genetic levels of nematode-trapping fungi. The results showed that microbial biogeography could be better understood by 1) using watersheds as research units, 2) removing the coverup of widespread species, and 3) applying good sampling efforts and strategies. We suggest that watersheds could help unify the understanding of the biogeographic patterns of animals, plants, and microbes and may also help account for the historical and contemporary factors driving species distributions.

Mesial temporal lobe epilepsy (mTLE) is one of the most common and refractory focal epilepsy syndromes. The molecular mechanisms of TLE are not completely understood. The aim of this study was to investigate the expression and potential function of plasma exosomal miRNAs (miR-483-5p, miR-671-5p, and miR-150-3p) in a mouse mode and in temporal lobe epilepsy patients. It was found that exosomal miRNAs were differentially expressed in three phases of the mouse mode, and exosomal miRNAs were down-regulated in mTLE patients compared with healthy controls. A bioinformatics analysis showed that target genes of exosomal miRNAs were significantly involved in the apoptotic process, cell adhesion, nervous system development, neurotrophin signaling pathway, PI3K-Akt signaling pathway, and metabolic pathways. The areas under the curve of miR-483-5p and miR-150-3p were 0.8714 (sensitivity = 75.00%, specificity = 91.65%) and 0.8213 (sensitivity = 67.50%, specificity = 90.00%), respectively. More importantly, the exosomal miRNAs were significantly associated with clinical parameters. Exosomal miRNAs may have the potential to become diagnostic and therapeutic biomarkers.