Foreground Capture Feature Pyramid Network-Oriented Object Detection in Complex Backgrounds.

Feature pyramids are widely adopted in visual detection models for capturing multiscale features of objects. However, the utilization of feature pyramids in practical object detection tasks is prone to complex background interference, resulting in suboptimal capture of discriminative multiscale foreground semantic features. In this article, a foreground capture feature pyramid network (FCFPN) for multiscale object detection is proposed, to address the problem of inadequate feature learning in complex backgrounds. FCFPN consists of a foreground dual attention (FDA) module and a pathway aggregation (PA) structure. Specifically, the FDA mechanism activates top-down foreground channel responses and lateral spatial foreground location features, so that channel and spatial foreground features are adequately captured. Then, the PA module adaptively learns the fusion weights of multiscale features at different levels of the feature pyramid, which enhances the complementarity of semantic information between different levels of the foreground feature maps. Since the fusion weights are learned adaptively based on different pyramid levels, the detection model accordingly retains the gained information of feature sizes and suppresses the conflicting information. The evaluations on public datasets and the self-built complex background dataset demonstrate that the detection average precision (AP) and the feature learning performance of the proposed method are superior compared with other FPNs, which proves the effectiveness of the proposed FCFPN.

Solar-driven CO2-to-ethanol conversion enabled by continuous CO2 transport via a superhydrophobic Cu2O nano fence.

The overall photocatalytic CO2 reduction reaction presents an eco-friendly approach for generating high-value products, specifically ethanol. However, ethanol production still faces efficiency issues (typically formation rates <605 µmol g-1 h-1). One significant challenge arises from the difficulty of continuously transporting CO2 to the catalyst surface, leading to inadequate gas reactant concentration at reactive sites. Here, we develop a mesoporous superhydrophobic Cu2O hollow structure (O-CHS) for efficient gas transport. O-CHS is designed to float on an aqueous solution and act as a nano fence, effectively impeding water infiltration into its inner space and enabling CO2 accumulation within. As CO2 is consumed at reactive sites, O-CHS serves as a gas transport channel and diffuser, continuously and promptly conveying CO2 from the gas phase to the reactive sites. This ensures a stable high CO2 concentration at reactive sites. Consequently, O-CHS achieves the highest recorded ethanol formation rate (996.18 µmol g-1 h-1) to the best of our knowledge. This strategy combines surface engineering with geometric modulation, providing a promising pathway for multi-carbon production.

Pulsed laser deposition of a Ga2O3 thin film for an optoelectronic synaptic device.

With the rapid development of information era, the traditional von Neumann architecture faces the computing bottleneck, and integration of memory and perception is regarded as a potential solution. Herein, a Ga2O3/Si heterojunction based multi-modulated optoelectronic synaptic device is fabricated and demonstrated. As stimulated by ultraviolet (UV) optical spikes, the heterojunction device reveals typical synaptic functions of excitatory-postsynaptic current (EPSC), paired-pulse facilitation (PPF), spike-timing-dependent plasticity (STDP), and switch between short-term memory (STM) and long-term memory (LTM). In addition, stronger stimulations like higher reading voltage, stronger optical stimulated intensity, and longer pulse duration time can significantly prolong the attenuation of EPSC, which contributes to the improvement of the forgetting process. Our work provides a potential strategy for future neuromorphic computation through a UV light driven stimulation.

First-principles study of magnetic interactions and excitations in antiferromagnetic van der Waals material MPX3(M=Mn, Fe, Co, Ni; X=S, Se).

Transition metal phosphorus trichalcogenides MPX3(M = Mn, Fe, Co, Ni; X = S, Se), as layered van der Waals antiferromagnetic (AFM) materials, have emerged as a promising platform for exploring two-dimensional (2D) magnetism. Based on density functional theory, we present a comprehensive investigation of the electronic and magnetic properties of MPX3. We calculated the spin exchange interactions as well as magnetocrystalline anisotropy energy. The numerical results reveal thatJ3is AFM in all cases, andJ2is significantly smaller compared to bothJ3andJ1. This behavior can be understood with regard to exchange paths and electron filling. Compared to other materials within this family, FePS3and CoPS3demonstrate significant easy-axis anisotropy. Using the obtained parameters, we estimated the Néel temperatureTNand Curie-Weiss temperatureθCW, and the results are in good agreement with the experimental observations. We further calculated the magnon spectra and successfully reproduce several typical features observed experimentally. Finally, we give helpful suggestions for the strong constraints about the range of non-negligible magnetic interactions based on the relations between magnon eigenvalues at high-symmetrykpoints in honeycomb lattices.

Spatial oblivion channel attention targeting intra-class diversity feature learning.

Convolutional neural networks (CNNs) have successfully driven many visual recognition tasks including image classification. However, when dealing with classification tasks with intra-class sample style diversity, the network tends to be disturbed by more diverse features, resulting in limited feature learning. In this article, a spatial oblivion channel attention (SOCA) for intra-class diversity feature learning is proposed. Specifically, SOCA performs spatial structure oblivion in a progressive regularization for each channel after convolution, so that the network is not restricted to a limited feature learning, and pays attention to more regionally detailed features. Further, SOCA reassigns channel weights in the progressively oblivious feature space from top to bottom along the channel direction, to ensure the network learns more image details in an orderly manner while not falling into feature redundancy. Experiments are conducted on the standard classification dataset CIFAR-10/100 and two garbage datasets with intra-class diverse styles. SOCA improves SqueezeNet, MobileNet, BN-VGG-19, Inception and ResNet-50 in classification accuracy by 1.31%, 1.18%, 1.57%, 2.09% and 2.27% on average, respectively. The feasibility and effectiveness of intra-class diversity feature learning in SOCA-enhanced networks are verified. Besides, the class activation map shows that more local detail feature regions are activated by adding the SOCA module, which also demonstrates the interpretability of the method for intra-class diversity feature learning.

Second-order Jahn-Teller effect induced high-temperature ferroelectricity in two-dimensional NbO2X (X = I, Br).

Based on the first-principles calculations, we investigated the ferroelectric properties of two-dimensional (2D) materials NbO2X (X = I, Br). Our cleavage energy analysis shows that exfoliating one NbO2I monolayer from its existing bulk counterpart is feasible. The phonon spectrum and molecular dynamics simulations confirm the dynamic and thermal stability of the monolayer structures for both NbO2I and NbO2Br. Total energy calculations show that the ferroelectric phase is the ground state for both materials, with the calculated in-plane ferroelectric polarizations being 384.5 pC m-1 and 375.2 pC m-1 for monolayers NbO2I and NbO2Br, respectively. Moreover, the intrinsic Curie temperature TC of monolayer NbO2I (NbO2Br) is as high as 1700 K (1500 K) from Monte Carlo simulation. Furthermore, with the orbital selective external potential method, the origin of ferroelectricity in NbO2X is revealed as the second-order Jahn-Teller effect. Our findings suggest that monolayers NbO2I and NbO2Br are promising candidate materials for practical ferroelectric applications.

Improving Biomedical Question Answering by Data Augmentation and Model Weighting.

Biomedical Question Answering aims to extract an answer to the given question from a biomedical context. Due to the strong professionalism of specific domain, it's more difficult to build large-scale datasets for specific domain question answering. Existing methods are limited by the lack of training data, and the performance is not as good as in open-domain settings, especially degrading when facing to the adversarial sample. We try to resolve the above issues. First, effective data augmentation strategies are adopted to improve the model training, including slide window, summarization and round-trip translation. Second, we propose a model weighting strategy for the final answer prediction in biomedical domain, which combines the advantage of two models, open-domain model QANet and BioBERT pre-trained in biomedical domain data. Finally, we give adversarial training to reinforce the robustness of the model. The public biomedical dataset collected from PubMed provided by BioASQ challenge is used to evaluate our approach. The results show that the model performance has been improved significantly compared to the single model and other models participated in BioASQ challenge. It can learn richer semantic expression from data augmentation and adversarial samples, which is beneficial to solve more complex question answering problems in biomedical domain.

Enhancement of Mass and Charge Transfer during Carbon Dioxide Photoreduction by Enhanced Surface Hydrophobicity without a Barrier Layer.

The CO2 reduction reaction (CRR) represents a promising route for the clean utilization of renewable resources. But mass-transfer limitations seriously hinder the forward step. Enhancing the surface hydrophobicity by using polymers has been proved to be one of the most efficient strategies. However, as macromolecular organics, polymers on the surface hinder the transfer of charge carriers from catalysts to reactants. Herein, we describe an in-situ surface fluorination strategy to enhance the surface hydrophobicity of TiO2 without a barrier layer of organics, thus facilitating the mass transfer of CO2 to catalysts and charge transfer. With less obstruction to charge transfer, a higher CO2, and lower H+ surface concentration, the photocatalytic CRR generation rate of methanol (CH3 OH) is greatly enhanced to up to 247.15 µmol g-1 h-1 . Furthermore, we investigated the overall defects; enhancing the surface hydrophobicity of catalysts provides a general and reliable method to improve the competitiveness of CRR.

UGDAS: Unsupervised graph-network based denoiser for abstractive summarization in biomedical domain.

Abstractive summarization models can generate summary auto-regressively, but the quality is often impacted by the noise in the text. Learning cross-sentence relations is a crucial step in this task and the graph-based network is more effective to capture the sentence relationship. Moreover, knowledge is very important to distinguish the noise of the text in special domain. A novel model structure called UGDAS is proposed in this paper, which combines a sentence-level denoiser based on an unsupervised graph-network and an auto-regressive generator. It utilizes domain knowledge and sentence position information to denoise the original text and further improve the quality of generated summaries. We use the recently-introduced dataset CORD-19 (COVID-19 Open Research Dataset) on text summarization task, which contains large-scale data on coronaviruses. The experimental results show that our model achieves the SOTA (state-of-the-art) result on CORD-19 dataset and outperforms the related baseline models on the PubMed Abstract dataset.

Graphene-mediated ferromagnetic coupling in the nickel nano-islands/graphene hybrid.

Nanoscale magnetic structures are fundamental to the design and fabrication of spintronic devices and have exhibited tremendous potential superior to the conventional semiconductor devices. However, most of the magnetic moments in nanostructures are unstable due to size effect, and the possible solution based on exchange coupling between nanomagnetism is still not clear. Here, graphene-mediated exchange coupling between nanomagnets is demonstrated by depositing discrete superparamagnetic Ni nano-islands on single-crystal graphene. The heterostructure exhibits ideal two-dimensional (2D) ferromagnetism with clear hysteresis loops and Curie temperature up to 80 K. The intrinsic ferromagnetism in graphene and antiferromagnetic exchange coupling between graphene and Ni nano-islands are revealed by x-ray magnetic circular dichroism and density functional theory calculations. The artificial 2D ferromagnets constitute a platform to study the coupling mechanism between complex correlated electronic systems and magnetism on the nanoscale, and the results and concept provide insights into the realization of spin manipulation in quantum computing.

Co-Immunization With CHIKV VLP and DNA Vaccines Induces a Promising Humoral Response in Mice.

Chikungunya fever is an acute infectious disease that is mediated by the mosquito-transmitted chikungunya virus (CHIKV), for which no licensed vaccines are currently available. Here, we explored several immunization protocols and investigated their immunity and protective effects in mice, with DNA- and virus-like particle (VLP)- vaccines, both alone and in combination. Both DNA and VLP vaccine candidates were developed and characterized, which express CHIKV structural genes (C-E3-E2-6K-E1). Mice were immunized twice, with different protocols, followed by immunological detection and CHIKV Ross challenge. The highest antigen-specific IgG and neutralizing activity were induced by DNA and VLP co-immunization, while the highest cellular immunity was induced by DNA vaccination alone. Although all vaccine groups could protect mice from lethal CHIKV challenge, demonstrated as reduced viral load in various tissues, without weight loss, mice co-immunized with DNA and VLP exhibited the mildest histopathological changes and lowest International Harmonization of Nomenclature and Diagnostic Criteria (INHAND) scores, in comparison to mice with either DNA or VLP vaccination alone. We concluded that co-immunization with DNA and VLP is a promising strategy to inducing better protective immunity against CHIKV infection.

Modulation on the Iron Centers by Selective Synthesis of Organic Ligands with Stereo-Specific Conformations.

Advanced fabrication of surface metal-organic complexes with specific coordination configuration and metal centers will facilitate to exploit novel nanomaterials with attractive electronic/magnetic properties. The precise on-surface synthesis provides an appealing strategy for in situ construction of complex organic ligands from simple precursors autonomously. In this paper, distinct organic ligands with stereo-specific conformation are separately synthesized through the well-known dehalogenative coupling. More interestingly, the exo-bent ligands promote the mono-iron chelated complexes with the Fe center significantly decoupled from the surface and of high spin, while the endo-bent ligands lead to bi-iron chelated ones instead with ferromagnetic properties.

Manipulating the Raman scattering rotation via magnetic field in an MoS2 monolayer.

Magneto-optical effects, which originate from the interactions between light and magnetism, have provided an important way to characterize magnetic materials and hosted abundant applications, such as light modulators, magnetic field sensors, and high-density data storage. However, such effects are too weak to be detected in non-magnetic materials due to the absence of spin degree of freedom. Here, we demonstrated that applying a perpendicular magnetic field can produce a colossal Raman scattering rotation in non-magnetic MoS2, for A-mode representing the out-of-plane breathing vibration. Our experimental results show that linearly polarized scattering light is rotated by ∓125°, more apparent than the valley Zeeman splitting effect (∓1.2 meV) under the same experimental conditions (±5 T), near room temperature. A detailed and systematic analysis on the polarization-resolved magnetic field-dependent micro-zone Raman intensity offers a feasible way to manipulate the inelastically scattered light via a magnetic technique. This explored phenomenology and physical mechanism arouse a new ramification of probing burgeoning magneto-optical effects in the field of two-dimensional laminar materials.

NS1-based DNA vaccination confers mouse protective immunity against ZIKV challenge.

The recent pandemic of Zika virus (ZIKV) infections highlight the urgent need for the development of a safe and efficacious ZIKV vaccine. We previously demonstrated that robust humoral and cellular immunity was elicited in BALB/c mice by ZIKV DNA vaccine encoding the precursor membrane (prM) and envelope (E) proteins while the protective efficacies were not evaluated against ZIKV challenge. To further explore the protective immunity elicited by various targets of ZIKV, we constructed a novel DNA-based vaccine expressing nonstructural protein 1 (NS1), named as VRC-NS1, and evaluated and compared immune responses and protective efficacies of three ZIKV DNA vaccine candidates (VRC-prME, VRC-NS1, and VRC-prME+VRC-NS1) using an A129 (Ifnar-/-) murine challenge model. The results showed that each of DNA vaccine candidates induced robust antigen-specific humoral immunity and conferred protection against ZIKV-SMGC-1 with two doses (20 µg per dose) of homologous intramuscularly (i.m.) immunizations via in vivo electroporation. All DNA vaccine candidates induced significant protection against infection-associated weight loss in addition to preventing viral replication in blood, brain and spleen tissue following in vivo viral challenge. Notably, NS1-based DNA vaccination alone was capable of conferring mouse protective immunity to reduce viremia and viral burden in tissues against ZIKV challenge, even though it did not induce neutralizing antibodies. These data demonstrated that VRC-NS1 and VRC-prME are highly promising vaccine candidates for ZIKV control. Furthermore, our results highlight an alternative strategy (DNA vaccine based on non-neutralizing antigen NS1) for designing novel flaviviral vaccines (including for ZIKV) and provide a foundation for the development of a safe and effective NS1-based vaccine against ZIKV infection.

Deep scaled dot-product attention based domain adaptation model for biomedical question answering.

Biomedical text mining is becoming increasingly important as the number of biomedical documents grow rapidly. Deep learning has boosted the development of biomedical text mining models. However, as deep learning models require a large amount of training data, a hierarchical attention based transfer learning model is proposed in this paper for the question answering task in biomedical field which lacks of sufficient training data. We adopt BERT (Bidirectional Encoder Representation Transformers), which has the ability to learn from large-scale unsupervised data, to enrich the semantic representation in our model. Especially, the scaled dot-product attention mechanism captures the question interaction clues for passage encoding. The domain adaptation technique of fine-tuning is used to reinforce the performance, which penalizes the deviations from the source model's parameters and remembers the knowledge of source domain. We evaluate the system performance on the open data set of BioASQ-Task B. The results show that our system achieves the state-of-the-art performance without any handcrafted features and outperforms the best solution for factoid questions in 2016 and 2017 BioASQ-Task B.

Magnetic Behavior in TiS3 Nanoribbon.

The electronic structure, magnetic properties and strain response of N-a-TiS3 nanoribbons are investigated by first-principles calculations. We find that the magnetic ground state is strongly dependent on width of a-TiS3. When N equals an odd number the ground state is a ferromagnetic (FM) metal, meanwhile, when N equals an even number the ground state is an anti-ferromagnetic (AFM) metal. More interestingly, a tensile strain as large as 6% can tune the 9-a-TiS3 nanoribbon from a FM metal to a half metal. A 4% tensile strain also causes a phase transition from AFM to FM ground state for 10-a-TiS3 nanoribbon. Our findings show that N-a-TiS3 is a promising candidate for spintronic and electronic applications.

Hydrogen Induced Etching Features of Wrinkled Graphene Domains.

Wrinkles are observed commonly in CVD (chemical vapor deposition)-grown graphene on Cu and hydrogen etching is of significant interest to understand the growth details, as well as a practical tool for fabricating functional graphene nanostructures. Here, we demonstrate a special hydrogen etching phenomenon of wrinkled graphene domains. We investigated the wrinkling of graphene domains under fast cooling conditions and the results indicated that wrinkles in the monolayer area formed more easily compared to the multilayer area (≥two layers), and the boundary of the multilayer area tended to be a high density wrinkle zone in those graphene domains, with a small portion of multilayer area in the center. Due to the site-selective adsorption of atomic hydrogen on wrinkled regions, the boundary of the multilayer area became a new initial point for the etching process, aside from the domain edge and random defect sites, as reported before, leading to the separation of the monolayer and multilayer area over time. A schematic model was drawn to illustrate how the etching of wrinkled graphene was generated and propagated. This work may provide valuable guidance for the design and growth of nanostructures based on wrinkled graphene.

Ultrahigh conductivity in Weyl semimetal NbAs nanobelts.

In two-dimensional (2D) systems, high mobility is typically achieved in low-carrier-density semiconductors and semimetals. Here, we discover that the nanobelts of Weyl semimetal NbAs maintain a high mobility even in the presence of a high sheet carrier density. We develop a growth scheme to synthesize single crystalline NbAs nanobelts with tunable Fermi levels. Owing to a large surface-to-bulk ratio, we argue that a 2D surface state gives rise to the high sheet carrier density, even though the bulk Fermi level is located near the Weyl nodes. A surface sheet conductance up to 5-100 S per □ is realized, exceeding that of conventional 2D electron gases, quasi-2D metal films, and topological insulator surface states. Corroborated by theory, we attribute the origin of the ultrahigh conductance to the disorder-tolerant Fermi arcs. The evidenced low-dissipation property of Fermi arcs has implications for both fundamental study and potential electronic applications.

Biomedical semantic indexing by deep neural network with multi-task learning.

BACKGROUND: Biomedical semantic indexing is important for information retrieval and many other research fields in bioinformatics. It annotates biomedical citations with Medical Subject Headings. In face of unbalanced category distribution in the training data, sampling methods are difficult to apply for semantic indexing task. RESULTS: In this paper, we present a novel deep serial multi-task learning model. The primary task treats the biomedical semantic indexing as a multi-label text classification issue that considers the relations of the labels. The auxiliary task is a regression task that predicts the MeSH number of the citation and provides hints for the network to make it converge faster. The experimental results on the BioASQ-Task5A open dataset show that our model outperforms the state-of-the-art solution "MTI", proposed by the US National Library of Medicine. Further, it not only achieves the highest precision among all the solutions in BioASQ-Task5A but also has faster convergence speed compared with some naive deep learning methods. CONCLUSIONS: Rather than parallel in an ordinary multi-task structure, the tasks in our model are serial and tightly coupled. It can achieve satisfied performance without any handcrafted feature.

A geographical location prediction method based on continuous time series Markov model.

Trajectory data uploaded by mobile devices is growing quickly. It represents the movement of an individual or a device based on the longitude and latitude coordinates collected by GPS. The location based service has a broad application prospect in the real world. As the traditional location prediction models which are based on the discrete state sequence cannot predict the locations in real time, we propose a Continuous Time Series Markov Model (CTS-MM) to solve this problem. The method takes the Gaussian Mixed Model (GMM) to simulate the posterior probability of a location in the continuous time series. The probability calculation method and state transition model of the Hidden Markov Model (HMM) are improved to get the precise location prediction. The experimental results on GeoLife data show that CTS-MM performs better for location prediction in exact minute than traditional location prediction models.