Self-attention (SA) network has shown profound value in image captioning. In this paper, we improve SA from two aspects to promote the performance of First, propose Normalized Self-Attention (NSA), a reparameterization that brings benefits normalization inside SA. While is previously only applied outside SA, introduce novel method and demonstrate it both possible beneficial perform on hidden activations Second, compensate for major limit Transformer fails model geometry structure input...

Deep convolutional neural network (DCNN) models have been widely explored for skin disease diagnosis and some of them achieved the diagnostic outcomes comparable or even superior to those dermatologists. However, broad implementation DCNN in detection is hindered by small size data imbalance publically accessible lesion datasets. This paper proposes a novel single-model based strategy classification lesions on imbalanced First, various DCNNs are trained different datasets verify that with...

In this paper, we propose an adversarial learning network for the task of multi-style image captioning (MSCap) with a standard factual caption dataset and multi-stylized language corpus without paired images. How to learn single model unpaired data is challenging necessary task, whereas rarely studied in previous works. The proposed framework mainly includes four contributive modules following typical encoder. First, style dependent generator output sentence conditioned on encoded specified...

. There is a need to develop high-performance and low-cost data augmentation strategies for intelligent skin cancer screening devices that can be deployed in rural or underdeveloped communities. The proposed strategy not only improve the classification performance of lesions but also highlight potential regions interest clinicians' attention. This implemented broad range clinical disciplines early automatic diagnosis many other diseases low resource settings.

Accurately manipulating the electronic structure of metal active sites under working conditions is central to developing efficient and stable electrocatalysts in industrial water-alkali electrolyzers. However, lack an intuitive means capture evolution during reaction state inhibits manipulation its structure. Here, atomically dispersed Ru single-sites on cobalt nanoparticles confined onto macro-microporous frameworks (M-Co NPs@Ru SAs/NC) with tunable electron coupling effect for catalysis...

Cervical abnormal cell detection plays a crucial role in the early screening of cervical cancer. In recent years, some deep learning-based methods have been proposed. However, these rely heavily on large amounts annotated images, which are time-consuming and laborintensive to acquire, thus limiting performance. this paper, we present novel Semi-supervised Abnormal Cell detector (SCAC), effectively utilizes abundant unlabeled data. We utilize Transformer as backbone SCAC capture long-range...

Salt pollution self-healing Al based solar evaporators are processed by using a picosecond laser. The evaporation rate can reach 2.325 kg m −2 h −1 with intensity of 0.95 kW .

The development of efficient and stable transition-metal-based electrocatalysts for the oxygen reduction reaction (ORR) in fuel cells is highly desirable, yet a great challenge remains. Here, we report novel Ni3N quantum dot (QD)/NiO heterostructure material, fabricated by immobilization metallic QDs onto surface NiO nanosheet, as active durable electrocatalyst performance. electrochemical characterizations theoretical calculations reveal that strong interface coupling effect QDs/NiO...

Developing low-cost yet efficient oxygen reduction reaction (ORR) electrocatalysts is of paramount importance for the widespread application renewable energy technologies. Here, a promising catalyst Co–Ni nanoalloy–organic framework (Co–Ni NOF) that significantly accelerates ORR kinetics process has been synthesized by straightforward pyrolysis strategy. This NOF with robust stability and strong methanol tolerance exhibits remarkable activity an excellent half-wave potential 0.88 V (30 mV...

Understanding the variation of active structure during hydrogen evolution reaction (HER) process is great importance for aiding in design optimized electrocatalysts. Herein, we present a composite material FeP nanoparticles coated by N-doped carbon (FeP@NC) as an efficient HER electrocatalyst, synthesized pyrolysis and equivalent-volume impregnation method. The as-prepared FeP@NC catalyst can accelerate at small overpotential 135 mV with current density 10 mA cm-2 acidic medium also shows...

Lack of intuitiveness and poor hand-eye coordination present a major technical challenge in neurosurgical navigation.We developed an integrated dexterous stereotactic co-axial projection imaging (sCPI) system featuring orthotopic image for augmented reality (AR) navigation. The performance characteristics the sCPI system, including resolution navigation accuracy, were quantitatively verified. was tested with USAF1951 test chart. accuracy measured using calibration panel 7×7 circle array...

A 3-D dynamic path-planning algorithm based on interfered fluid flow (DA) is proposed for unmanned aerial vehicle (UAV) in environment. This paper first describes the mathematical modeling of convex obstacles. Then static (SA) introduced and improved, whose planning path conforms to general characteristic phenomenon that running water can avoid rock arrive at destination. By updating information obstacles obtaining relative velocity, we then transform problem problem. Therefore, DA...

Self-attention (SA) network has shown profound value in image captioning. In this paper, we improve SA from two aspects to promote the performance of First, propose Normalized Self-Attention (NSA), a reparameterization that brings benefits normalization inside SA. While is previously only applied outside SA, introduce novel method and demonstrate it both possible beneficial perform on hidden activations Second, compensate for major limit Transformer fails model geometry structure input...

Although deep convolutional neural networks (DCNNs) have achieved significant accuracy in skin lesion classification comparable or even superior to those of dermatologists, practical implementation these models for cancer screening low resource settings is hindered by their limitations computational cost and training dataset. To overcome limitations, we propose a low-cost high-performance data augmentation strategy that includes two consecutive stages search network search. At the stage,...

Autonomous obstacle avoidance technology is the key to determine whether autonomous underwater vehicle (AUV) can reach its destination safely. The Q-Learning based guidance vector field proposed solve 3-dimensional problem for AUV. Firstly initial in free space constructed guide AUV along shortest path. Then modulation matrix quantify influence generated by obstacles so that modified environment obtained. For case of entering trap area, Q-learning algorithm used find path from current...

Using networking and embedded technology, electricity management system of ternary equipment two-tier network, which is composed host control computer, sale locale cabinet power meter to form around campus RS-485 bus realized the intelligent administration apartment with security, economy.

<p>To accurately and efficiently capture the topological attribute information of nodes apply them to link prediction task, this paper proposes a Dual Channel Graph Convolution Link Prediction (DC-GCN). DC-GCN constructs dual channel through graph convolution network. can learn both embeddings nodes; it introduces an attention mechanism weights each embedding adaptively then performs weighted fusion obtain final representation nodes. Finally, Hadamard distance is used construct between...

Designing neural architectures requires immense manual efforts. This has promoted the development of architecture search (NAS) to automate design. While previous NAS methods achieve promising results but run slowly, zero-cost proxies extremely fast are less promising. Therefore, it is great potential accelerate via those proxies. The existing method two limitations, which unforeseeable reliability and one-shot usage. To address we present ProxyBO, an efficient Bayesian optimization (BO)...

Image captioning aims to first observe an image, most notably the involved objects that are highly context-dependent, and then depict it with a natural description. However, of current models solely use isolated vectors as image representations, ignoring contexts among them. In this paper, we introduce Local-Global Context (LGC) network, endowing independent object features shortrange perception (local contexts) long-range dependence (global contexts). LGC network can be viewed feature...