Search

In the rapidly advancing field of Underwater Acoustic Communication (UAC), channel prediction remains a major challenge, exacerbated by the complicated nature of ocean environments. This paper introduces an innovative Multi- Task Learning (MTL) framework for time-varying Underwater Acoustic (UWA) channel prediction. By decomposing the highdimensional Channel Impulse Response (CIR) prediction into interconnected tasks, the proposed framework leverages a Shared Feature Learning (SFL) layer, capturing intricate dependencies underlying UWA channels. To validate its efficacy, we conducted thorough evaluations, leveraging real-world data from two distinct at-sea experiments conducted in Wuyuan Bay, China. A comprehensive comparative study of various configurations for the SFL layer, ranging from commonly used Recurrent Neural Network (RNN)-based models to the more advanced transformer structure, further underscores the flexibility and broad applicability of our MTL framework for handling various challenging UWA environments.

In the training procedures of many real-world learning models, gathering and annotating decent amounts of labeled data can be cost-prohibitive. To mitigate this data-hungry problem, active learning (AL) and semi-supervised learning (SSL) are frequently adopted as two effective but often isolated means. Some recent studies explored the potential of combining AL and SSL to better probe the unlabeled data. However, almost all these contemporary SSL-AL works use a simple combination strategy, ignoring SSL and AL's inherent relation. Further, other methods suffer from high computational costs when dealing with large-scale, high-dimensional datasets. Motivated by the industry practice of labeling data, we first propose an innovative Inconsistency-based virtual aDvErsarial Active Learning (IDEAL) algorithm to further investigate SSL-AL's potential superiority and achieve mutual enhancement of AL and SSL, i.e., SSL propagates label information to unlabeled samples and provides smoothed embeddings for AL, while AL excludes samples with inconsistent predictions and considerable uncertainty for SSL. We estimate unlabeled samples’ inconsistency by augmentation strategies of different granularities, including fine-grained continuous perturbation exploration and coarse-grained data transformations. Moreover, to solve the problem that the utilization efficiency of unlabeled samples is still insufficient in the process of semi-supervised training, we extend our IDEAL to a curriculum-guided version, namely SPL-IDEAL algorithm. The SPL-IDEAL algorithm can regularize the training process towards better regions in parameter space and denoise the pseudo labels with low confidence, achieving better performance. The extensive experiments, in both text and image benchmark datasets, validate the effectiveness of our proposed IDEAL and SPL-IDEAL algorithms, comparing them against state-of-the-art baselines. Two real-world case studies visualize the practical industrial value of applying and deploying the proposed data sampling algorithms.

Zoned Namespace (ZNS) SSDs present a new class of storage devices that promise low cost, stable performance, and software-defined capability. ZNS abstracts the SSD into an array of zones that can only be written sequentially. Although ZNS-compatible filesystems (e.g., F2FS) launch sequential writes, the zone write constraint may be violated due to orderless Linux I/O stack. Hence, the write queue depth of each zone must be limited to one, which severely degrades the performance of small writes. In this paper, we propose oZNS SSD (o: orderless), which allows multiple writes to be submitted to a zone and processed out-of-order in the SSD. To bridge the gap between ZNS sequential write contract and orderless writes on flash, a lightweight indirection layer is introduced and carefully designed by exploiting the characteristics of out-of-order writes. Specifically, memory-efficient metadata structures are devised to record the write deviations of data pages, and a two-tier buffering mechanism is employed to reduce metadata and accelerate metadata access. Moreover, a read-try policy removes metadata access from read critical path and thus improves the data read efficiency. Experimental results show that oZNS can achieve up to 8.6$\times$× higher write performance than traditional ZNS while providing comparable read performance.

Implicit large-eddy simulation of an over-expanded screeching rectangular jet is performed with a seventh-order finite difference scheme. Good agreement is found between the predicted flow- and acoustic fields with the experimental observations. Fourier decomposition, phase-averaging analysis and Spectral Proper Orthogonal Decomposition (SPOD) are used to investigate the origin of the screech, the shock leakage during the shear-layer flapping, and the distinguishing fluctuating characteristics in the minor- and major-axis plane of the rectangular jet. It finds that the screech is radiated from the end of the forth shock cell, where the interaction of the shock waves with the shear layer causes periodic leakages of shock-wave tips in the minor-axis plane, resulting in the generation of intense acoustic waves in the surrounding air. An obvious flapping mode at the same frequency of the screech is captured in the minor-axis plane and dominates the dynamic motions of the rectangular jet. The SPOD modes of pressure and velocity fluctuations at the screech frequency help to reveal the relationship between the screech generation and the coherent structures.

In this article, the design, fabrication, and comparison of Pb(Mg1/3 Nb2/3) O3-0.25PbTiO3 (PMN-0.25PT) single crystals (SCs) by alternating current poling (ACP) and direct current poling (DCP)-based 2.8-MHz phased array ultrasonic transducers were investigated. Pzflex finite element simulation was used to model and compare the electro-acoustic properties of ACP and DCP phased arrays. The finite element analysis (FEA) results indicated that the ACP phased arrays can exhibit a greater sensitivity, a broader -6-dB bandwidth (BW), and a lower center frequency under the same acoustic structure and the same thickness of matching layers. In addition, the high coincidence of simulated and measured impedance proves the stability of the transducer fabrication processes. By comparing the permittivity of the SCs in the processes of transducer fabrication, it is shown that the ACP SCs are more sensitive to the clamping force during the pressurization process and to the cutting force during the dicing process, respectively. Compared with the DCP phased arrays, the ACP phased arrays have been shown to exhibit 7.90% increase in observed -6-dB BW and 0.82-dB increase in relative sensitivity (RS). These results suggest that ACP phased arrays can enhance both the longitudinal resolution and penetration depth of ultrasonic imaging. Finally, a typical phantom image is created to compare the imaging capabilities of ACP and DCP phased arrays. The findings indicate that phased arrays based on ACP PMNT SCs might be promising for medical ultrasound (US) imaging applications.

Transformers have emerged as the leading methods in natural language processing, computer vision, and multi-modal applications due to their ability to capture complex relationships and dependencies in data. In this study, we explore the potential of transformers as feature aggregators in the context of patch-based writer retrieval, with the objective of improving the quality of writer retrieval by effectively summarizing the relevant features from image patches. Our investigation underscores the complexity of leveraging transformers as feature aggregators in patch-based writer retrieval. While we have experimented with various model configurations, augmentations, and learning objectives, the performance of transformers in this task has room for improvement. This observation highlights the challenges in this domain and emphasizes the need for further research to enhance their effectiveness. By shedding light on the limitations of transformers in this context, our study contributes to the growing body of knowledge in the field of writer retrieval and provides valuable insights for future research and development in this area.

Emerging high-performance computing (HPC) applications with diverse workload characteristics impose greater demands on parallel file systems (PFSs). PFSs also require more efficient software designs to fully utilize the performance of modern hardware, such as multi-core CPUs, Remote Direct Memory Access (RDMA), and NVMe SSDs. However, existing PFSs expose great limitations under these requirements due to limited multi-core scalability, unaware of HPC workloads, and disjointed network-storage optimizations. In this article, we present PeakFS, an ultra-high performance parallel file system via computing-network-storage co-optimization for HPC applications. PeakFS designs a shared-nothing scheduling system based on link-reduced task dispatching with lock-free queues to reduce concurrency overhead. Besides, PeakFS improves I/O performance with flexible distribution strategies, memory-efficient indexing, and metadata caching according to HPC I/O characteristics. Finally, PeakFS shortens the critical path of request processing through network-storage co-optimizations. Experimental results show that the metadata and data performance of PeakFS reaches more than 90% of the hardware limits. For metadata throughput, PeakFS achieves a 3.5–19× improvement over GekkoFS and outperforms BeeGFS by three orders of magnitude.

Unveiling molecular mechanisms that dominate protein phase dynamics has been a pressing need for deciphering the intricate intracellular modulation machinery. While ions and biomacromolecules have been widely recognized for modulating protein phase separations, effects of small molecules that essentially constitute the cytosolic chemical atmosphere on the protein phase behaviors are rarely understood. Herein, we report that vitamin C (VC), a key small molecule for maintaining a reductive intracellular atmosphere, drives reentrant phase transitions of myosin II/F-actin (actomyosin) cytoskeletons. The actomyosin bundle condensates dissemble in the low-VCregime and assemble in the high-VCregime in vitro or inside neuronal cells, through a concurrent myosin II protein aggregation–dissociation process with monotonic VCconcentration increase. Based on this finding, we employ in situ single-cell and single-vesicle electrochemistry to demonstrate the quantitative modulation of catecholamine transmitter vesicle exocytosis by intracellular VCatmosphere, i.e., exocytotic release amount increases in the low-VCregime and decreases in the high-VCregime. Furthermore, we show how VCregulates cytomembrane-vesicle fusion pore dynamics through counteractive or synergistic effects of actomyosin phase transitions and the intracellular free calcium level on membrane tensions. Our work uncovers the small molecule-based reversive protein phase regulatory mechanism, paving a new way to chemical neuromodulation and therapeutic repertoire expansion.

Entity relationship extraction is the core task of text mining and intelligent retrieval, which can automatically extract the semantic relationships between entities. For a problem of low accuracy of information extraction due to semantic complexity and nested entities in the text in the field of electric power communication network, this paper proposes an Orth-Biaff-CasRel method for joint extraction of entity relationships in faulty text. The method incorporates the RoFormer of rotational position as a coding layer in the encoding process, capturing the relative positional relationships between entities in the text; In addition, considers the entity content information and entity boundary information, designs the head entity extraction method with Orthogonalized Biaffine attention mechanism. Finally, combining stacked pointers and adding hidden layers in the joint extraction of tail entities and relationships to achieve accurate extraction of information from complex semantic fault texts. The proposed method is experimentally verified in real fault texts of power communication networks and achieves good results, significantly outperforming the existing methods in terms of accuracy, recall and Fl value, with an improvement of 4.41 %,1.43 % and 2.92 %, respectively.

Polyamides represent one class of materials that is important in modern society. Because of the numerous potential applications of polyamides in various fields, there is a high demand for new polyamide structures, which necessitates the development of new polymerization methods. Herein, we report a novel and efficient palladium-catalyzed hydroaminocarbonylative polymerization of dienes and diamines for the synthesis of cycloaliphatic polyamides. The method employs readily available starting materials, proceeds in an atom-economic manner, and creates a series of new functional polyamides in high yields and high molecular weights. In contrast with the traditional polyamides based on adipic acid, the cycloaliphatic polyamides have superior thermal resistance, higher glass-transition temperature, and better solubility in common organic solvents, thus probably featuring the merits of high-performance and good processability.

Developing catalysts with high catalytic activity and stability in acidic media is crucial for advancing hydrogen production in proton exchange membrane water electrolyzers (PEMWEs). To this end, a self-supported WO3@RuO2nanowire structure was grown in situ on a titanium mesh using hydrothermal and ion-exchange methods. Despite a Ru loading of only 0.098 wt %, it achieves an overpotential of 246 mV for the oxygen evolution reaction (OER) at a current density of 10 mA·cm–2in acidic 0.5 M H2SO4while maintaining excellent stability over 50 h, much better than that of the commercial RuO2. After the establishment of the WO3@RuO2heterostructure, a reduced overpotential of the rate-determining step from M–O* to M–OOH* is confirmed by the DFT calculation. Meanwhile, its enhanced OER kinetics are also greatly improved by this self-supported system in the absence of the organic binder, leading to a reduced interface resistance between active sites and electrolytes. This work presents a promising approach to minimize the use of noble metals for large-scale PEMWE applications.

Tannic acid (TA) forms a cross-linked network on the surface of perovskite, which can effectively passivate defects and inhibit I−ions diffusion.

Fractal theory offers a powerful tool for the precise description and quantification of the complex pore structures in reservoir rocks, crucial for understanding the storage and migration characteristics of media within these rocks. Faced with the challenge of calculating the three-dimensional fractal dimensions of rock porosity, this study proposes an innovative computational process that directly calculates the three-dimensional fractal dimensions from a geometric perspective. By employing a composite denoising approach that integrates Fourier transform (FT) and wavelet transform (WT), coupled with multimodal pore extraction techniques such as threshold segmentation, top-hat transformation, and membrane enhancement, we successfully crafted accurate digital rock models. The improved box-counting method was then applied to analyze the voxel data of these digital rocks, accurately calculating the fractal dimensions of the rock pore distribution. Further numerical simulations of permeability experiments were conducted to explore the physical correlations between the rock pore fractal dimensions, porosity, and absolute permeability. The results reveal that rocks with higher fractal dimensions exhibit more complex pore connectivity pathways and a wider, more uneven pore distribution, suggesting that the ideal rock samples should possess lower fractal dimensions and higher effective porosity rates to achieve optimal fluid transmission properties. The methodology and conclusions of this study provide new tools and insights for the quantitative analysis of complex pores in rocks and contribute to the exploration of the fractal transport properties of media within rocks.

ABSTRACTMyogenic differentiation (MyoD) 1, which is known as a pivotal transcription factor during myogenesis, has been proven dysregulated in several cancers. However, litter is known about the precise role and downstream genes of MyoD1 in gastric cancer (GC) cells. Here, we report that MyoD1 is lowly expressed in primary GC tissues and cells. In our experiments, overexpression of MyoD1 inhibited cell proliferation. Downstream genes of MyoD1 regulation were investigated using RNA-Seq. As a result, 138 up-regulated genes and 20 down-regulated genes and 27 up-regulated lncRNAs and 20 down-regulated lncRNAs were identified in MyoD1 overexpressed MKN-45 cells, which participated in epithelial cell signaling in Helicobacter pylori infection, glycosaminoglycan biosynthesis (keratan sulfate), notch signaling pathway, and others. Among these genes, BIK was directly regulated by MyoD1 in GC cells and inhibited cancer cell proliferation. The BIK knockdown rescued the effects of MyoD1 overexpression on GC cells. In conclusion, MyoD1 inhibited cell proliferation via 158 genes and 47 lncRNAs downstream directly or indirectly that participated in multiple signaling pathways in GC, and among these, MyoD1 promotes BIK transcription by binding to its promoter, then promotes BIK-Bcl2-caspase 3 axis and regulates GC cell apoptosis.

Acute ischemic stroke followed by microglia activation, and the regulation of neuroinflammatory responses after ischemic injury involves microglia polarization. microglia polarization is involved in the regulation of neuroinflammatory responses and ischemic stroke-related brain damage. Thymoquinone (TQ) is an anti-inflammatory agent following ischemic stroke onset. However, the significance of TQ in microglia polarization following acute ischemic stroke is still unclear. We predicted that TQ might have neuroprotective properties by modulating microglia polarization. In this work, we mimicked the clinical signs of acute ischemic stroke using a mouse middle cerebral artery ischemia-reperfusion (I/R) model. It was discovered that TQ treatment decreased I/R-induced infarct volume, cerebral oedema, and promoted neuronal survival, as well as improved the histopathological changes of brain tissue. The sensorimotor function was assessed by the Garica score, foot fault test, and corner test, and it was found that TQ could improve the motor deficits caused by I/R. Secondly, real-time fluorescence quantitative PCR, immuno-fluorescence, ELISA, and western blot were used to detect the expression of M1/M2-specific markers in microglia to explore the role of TQ in the modulation of microglial cell polarization after cerebral ischemia-reperfusion. We found that TQ was able to promote the polarization of microglia with extremely secreted inflammatory factors from M1 type to M2 type. Furthermore, TQ could block the TLR4/NF-κB signaling pathway via Hif-1α activation which subsequently may attenuate microglia differentiation following the cerebral ischemia, establishing a mechanism for the TQ's beneficial effects in the cerebral ischemia-reperfusion model.

Recommendation performance usually exhibits a long-tail distribution over users — a small portion of head users enjoy much more accurate recommendation services than the others. We reveal two sources of this performance heterogeneity problem: the uneven distribution of historical interactions (a natural source); and the biased training of recommender models (a model source). As addressing this problem cannot sacrifice the overall performance, a wise choice is to eliminate the model bias while maintaining the natural heterogeneity. The key to debiased training lies in eliminating the effect of confounders that influence both the user's historical behaviors and the next behavior. The emerging causal recommendation methods achieve this by modeling the causal effect between user behaviors, however potentially neglect unobserved confounders (e.g., friend suggestions) that are hard to measure in practice. To address unobserved confounders, we resort to the front-door adjustment (FDA) in causal theory and propose a causal multi-teacher distillation framework (CausalD). FDA requires proper mediators in order to estimate the causal effects of historical behaviors on the next behavior. To achieve this, we equip CausalD with multiple heterogeneous recommendation models to model the mediator distribution. Then, the causal effect estimated by FDA is the expectation of recommendation prediction over the mediator distribution and the prior distribution of historical behaviors, which is technically achieved by multi-teacher ensemble. To pursue efficient inference, CausalD further distills multiple teachers into one student model to directly infer the causal effect for making recommendations. We instantiate CausalD on two representative models, DeepFM and DIN, and conduct extensive experiments on three real-world datasets, which validate the superiority of CausalD over state-of-the-art methods. Through in-depth analysis, we find that CausalD largely improves the performance of tail users, reduces the performance heterogeneity, and enhances the overall performance.

Dependency-based models are widely used to extract semantic relations in text. Most existing dependency-based models establish stacked structures to merge contextual and dependency information, which encode the contextual information first and then encode the dependency information. However, this unidirectional information flow weakens the representation of words in the sentence, which further restricts the performance of existing models. To establish bidirectional information flow, a dual attention graph convolutional network (DAGCN) with a parallel structure is proposed. Most importantly, DAGCN can build multi-turn interactions between contextual and dependency information to imitate the multi-turn looking-back actions of human beings. In addition, multi-layer adjacency matrix-aware multi-head attention (AMAtt), including context-to-dependency attention and dependency-to-context attention, is carefully designed as a merge mechanism in the parallel structure to preserve the structural information of sentences and dependency trees during interactions. Furthermore, DAGCN is evaluated on the popular PubMed dataset, TACRED dataset and SemEval 2010 Task 8 dataset to demonstrate its validity. Experimental results show that our model outperforms the existing dependency-based models.

In the realm of medical diagnosis, recent strides in deep neural network-guided magnetic resonance imaging (MRI) restoration have shown promise. Nevertheless, persistent drawbacks overshadow these advancements. Challenges persist in balancing acquisition speed and image quality, while existing methods primarily focus on singular tasks like MRI reconstruction or super-resolution (SR), neglecting the interplay between these tasks. To tackle these challenges, this paper introduces the mutual co-attention network (MCAN) specifically designed to concurrently address both MRI reconstruction and SR tasks. Comprising multiple mutual cooperation attention blocks (MCABs) in succession, MCAN is tailored to maintain consistency between local physiological details and global anatomical structures. The intricately crafted MCAB includes a feature extraction block, a local attention block and a global attention block. Additionally, to ensure data fidelity without compromising acquired data, we propose the channel-wise data consistency block. Thorough experimentation on the IXI and fastMRI dataset showcases MCAN’s superiority over existing state-of-the-art methods. Both quantitative metrics and visual quality assessments validate the enhanced performance of MCAN in MRI restoration. The findings underscore MCAN’s potential in significantly advancing therapeutic applications. By mitigating the trade-off between acquisition speed and image quality while simultaneously addressing both MRI reconstruction and SR tasks, MCAN emerges as a promising solution in the domain of magnetic resonance image restoration.Graphical Abstract

To solve the problems of low strength and unsatisfactory connection quality of clinching in three-layer sheets, a method of clinching aluminum alloy three-layer sheets with an inclined wall die was proposed. AA5052 aluminum alloy sheets were chosen as the research objects. Elastic-plastic finite element simulation and experimental research were conducted to analyze the mechanism of clinching joints in three-layer sheets and a finite element method has been used to investigate and summarize the five stages of forming a three-layer aluminum sheet clinching. Six load distribution methods were designed to investigate the strength of clinching joints in three-layer sheets. The failure mode and failure mechanism of the joints in the shear and peel experiment were comprehensively studied. The comparison between the experimental and finite element results had an error of less than 8%, indicating that the finite element model was highly reliable. The failure strength of the joints and the energy absorption values were determined by physical experiments, and the results show that the D-2 joint has a maximum shear strength of 3400 N and an energy absorption of 6.67 J. The D-3 joint has a shear strength of 3040 N and an energy absorption of 9.29 J. The T joint has a lower strength and a higher failure displacement, with the T-3 joint having a maximum peel strength of 690 N and an energy absorption of 11.876 J.

Light-detection-and-ranging-based multiobject detection and tracking play fundamental roles in autonomous driving systems. Most existing detection and tracking methods inevitably require complex pairing permutations for object association across frames, making the framework slow. Moreover, the occlusion and viewpoint changes lead to missed and false detection. To solve the abovementioned issues, this article proposes a simple joint 3-D detection and tracking approach with three-step offset learning ($\text{Offset}^{3}\text{Net}$). Specifically, $\text{Offset}^{3}\text{Net}$ incorporates three task-specific output subnetworks to learn three offsets: 1) center offset, 2) motion offset, and 3) association offset. The learning of abovementioned offsets eliminates the complex bipartite matching processing. Specifically, the center offset guides the model to generate precise detections, whereas the motion offset transforms the track from the previous frame to the current frame, and the association offset minimizes the distance between detection and motion-updated track of the same object. Then, a simple read-off operation is conducted for data association on a hybrid-time centerness map, which represents the detections and offset-updated tracks. In addition, we design a detection-feature-enhanced module that captures the temporal coherence of the object motion and appearance information, avoiding the missed and false detection. Experiments on nuScenes have demonstrated the effectiveness of our $\text{Offset}^{3}\text{Net}$ in terms of accuracy and speed compared with most 3-D detection and tracking methods.

As an emerging technique, deep unfolding networks (DUNs) has drawn increasing attentions in the field of MRI reconstruction. With the integration of traditional iterative algorithms and modern learning-based blocks, DUNs enjoy extensive reputations of high interpretability and easy tractability. However, the performance of most DUNs can hardly reach a clinical level due to several limitations such as cross-stage independence, feature insufficiency, and interaction deficiency. To address these issues, we propose a novel information-growth holistic unfolding network, dubbed IHUN, which is stage-wisely armed with two main steps, i.e., a coarse estimation to guarantee data consistency and a feature calibration to suppress noises. More importantly, three collaborated subassemblies, including information-growth feature fusion block (I2FB), holistic feature extraction block (HFEB), and high-frequency cross feature fusion block (H3FB), are used in the feature calibration step to ensure a desirable information flow. Specifically, I2FB hierarchically aggregates the hidden states of previous stages to effectively update current solution. Operating in a computationally efficient fashion, HFEB takes charge of capturing the local/global features and boosting the inductive capacity. H3FB effectively fuses high-frequency information from two distinct modalities, thereby enabling the precise restoration of intricate details. Extensive experiments conducted on benchmark MRI datasets demonstrate the appealing performance of the proposed method, in terms of both quality and quantity. Codes are available at (https://github.com/chenjiachengzzz/IHUN).

The one-shot multiple object tracking (MOT) framework simultaneously generates detected targets and re-identification (ReID) embeddings, and employs them to associate previous tracks. However, the continuity between the learning and extraction within the ReID task is unconsciously neglected because of treating them as two isolated stages. This leads to unreliable ReID embeddings and poses a challenge to target matching especially in complex and crowded scenarios. To this end, we propose a guided embedding enhancement multiple object tracker, named G2EMOT. It consists of three innovative designs aiming at improving the embedding representation capability for individual targets. First, based on the original feature map, a global instance-specific context decoupling (GICD) module is devised to facilitate the respective feature learning for detection and ReID branches. Then, a heatmap-guided embedding enhancement (HGEE) module is introduced to connect the processes of embedding learning and extraction, ensuring that the detected target coordinates are accurately aligned with the discriminative ReID embeddings. Finally, to associate and stabilize matched targets, we present a novel two-gate guided embedding update (TG-GEU) strategy that dynamically updates ReID embeddings. With the proposed three components, G2EMOT achieves outstanding performance on popular MOT benchmarks while outperforming the existing one-shot tracking methods by a large margin. In particular, it realizes IDF1 of 76.1% on MOT17 and IDF1 of 74.2% on MOT20. The source codes are released at https://github.com/ydhcg-BoBo/G2EMOT.

Recently, multi-view feature learning technique has attracted lots of research interest. Discriminant analysis-based multi-view feature learning is an important research branch. Although some multi-view discriminant analysis methods have been presented, there still exists room for improvement. How to effectively explore the discriminant and local geometrical structure information simultaneously from multiple views is still an important research topic. In this paper, we propose a novel approach named uncorrelated locality-sensitive multi-view discriminant analysis, which jointly learns multiple view-specific transformations, such that in the projected subspace for each view, the within-class nearby samples are close to each other, while the between-class nearby samples are far apart. We provide a multi-view sample distance term to promote the one-to-one data consistency across views. Furthermore, we design uncorrelated constraints to reduce the redundancy among the transformations. Experiments on two widely used datasets demonstrate the effectiveness of the proposed approach.

Cross-view geo-localization has been widely used as an important technique for determining the geographical location of unmanned aerial vehicles (UAVs). Despite various image retrieval methods proposed, drone and satellite image cross-view geo-localization still remain challenging due to their wildly inconsistent view angles. In this article, we propose a new framework, the Swin-radial-locality network (SRLN), to extract robust image feature representations. Specifically, SRLN is based on a pruned version of the Swin transformer, which integrates multiscale feature aggregation within a Siamese network structure, featuring shared weights and equipped with multiclassification heads. SRLN is mainly comprised of a radial-slicer-network (RSN) and a local-pattern-network (LPN), which is designed to effectively harmonize directional information from drone-captured images and broader environmental features from satellite imagery, crucial for capturing angle and feature details between drone and satellite images. The RSN part focuses on capturing fine-grained features that represent the drone’s directional information, while the LPN is utilized for a more comprehensive analysis of broader environmental features. Extensive experiments are carried out on widely used public benchmark datasets, i.e., University-1652 and SUES-200. With more than 3% improvement over existing methods in both drone-view target localization tasks and drone navigation applications, the results validate the superior performance of our multiscale feature fusion model, achieving a state-of-the-art performance record.

Video moment retrieval (VMR) aims to localize the target moment in an untrimmed video according to the given nature language query. The existing algorithms typically rely on clean annotations to train their models. However, making annotations by human labors may introduce much noise. Thus, the video moment retrieval models will not be well trained in practice. In this article, we present a simple yet effective video moment retrieval framework via bottom-up schema, which is in end-to-end manners and robust to noisy label training. Specifically, we extract the multimodal features by syntactic graph convolutional networks and multihead attention layers, which are fused by the cross gates and the bilinear approach. Then, the feature pyramid networks are constructed to encode plentiful scene relationships and capture high semantics. Furthermore, to mitigate the effects of noisy annotations, we devise the multilevel losses characterized by two levels: a frame-level loss that improves noise tolerance and an instance-level loss that reduces adverse effects of negative instances. For the frame level, we adopt the Gaussian smoothing to regard noisy labels as soft labels through the partial fitting. For the instance level, we exploit a pair of structurally identical models to let them teach each other during iterations. This leads to our proposed robust video moment retrieval model, which experimentally and significantly outperforms the state-of-the-art approaches on standard public datasets ActivityCaption and textually annotated cooking scene (TACoS). We also evaluate the proposed approach on the different manual annotation noises to further demonstrate the effectiveness of our model.

Acrokeratoelastoidosis (AKE) is a rare autosomal dominant hereditary skin disease characterized by small, round-oval, flat-topped keratotic papules on the palms, soles and dorsal aspect of hands or feet. The causative gene for AKE remains unidentified. This study aims to identify the causative gene of AKE and explore the underlying biological mechanisms. A large, three-generation Chinese family exhibiting classic AKE symptoms was identified. A genome-wide linkage analysis and whole-exome sequencing were employed to determine the causative gene. shRNA knockdown in human skin fibroblasts and CRISPR/Cas9 knockout in HEK293T cells were utilized to assess gene functions in the progression of elastic fiber biosynthesis. The linkage analysis identified a susceptibility region between rs7296765 to rs10784618 on chromosome 12. Whole-exome sequencing confirmed a splicing mutation of 1101 + 1 G> Ain the CCDC91gene, resulting in exon 11 skipping and a subsequent 59-amino-acid-residue loss (residues L309-Q367del). Further functional analysis revealed distended Golgi cisternae, cytoplasmic vesicle accumulation, and lysosome presence. Immnunostaining of si-CCDC91-HSF cells demonstrated tropoelastin accumulation in the Golgi and abnormal extracellular aggregates. There are no significant changes in Fibrillin-1 microfibril assembly and lysyl oxidase activity. The findings strongly suggest that the protein product of the CCDC91gene plays a crucial role in elastin transport. This discovery enhances our understanding of CCDC91’s function and broadens the known pathogenic mechanisms of AKE.

Memory-semantic solid-state drives (MS-SSDs) provide a promising opportunity to enable the hybrid memory architecture (HMA). The memory-semantic interface enables the CPUs to directly access structured data in SSDs and eliminate bulk data copy/swap between the memory and storage devices. However, existing hash indexings issue many random writes, resulting in two problems when directly deployed on MS-SSD-based HMA: 1) highly random traffic persisted to the underlying nand flash of MS-SSDs incurs significant garbage collection (GC) overhead and 2) placing frequently updated memory pages of hash indexings in persistent memories (PMs) is anticipated to reduce write latency, failing to work effectively due to the lack of skewness. To address the above problems, we propose a novel MS-SSD-friendly hash indexing scheme called TieredHM. It employs a multilayer structure and opportunistic data movement (ODM) to construct skewed writes. Hence, the MS-SSD can transform the writes into multistreamed writes, separating data with different update frequencies to reduce GC overhead. Besides, since the top layer is updated much more frequently (more skewed) than other layers, placing the top layer of TieredHM into PM can significantly reduce write latency. TieredHM further leverages a prefetch mechanism based on the internal parallelism of nand flash to reduce search overhead incurred by ODM. Experimental results show that TieredHM reduces the average write latency and GC overhead by up to $8.3 \times $ and $20.0 \times $ compared to state-of-the-art hash indexings without sacrificing read performance.

The integration of predictive, preventive, personalized, and participatory (P4) healthcare advocates proactive intervention, including dietary supplements and lifestyle interventions for chronic disease. Personal profiles include deep phenotypic data and genetic information, which are associated with chronic diseases, can guide proactive intervention. However, little is known about how to design an appropriate intervention mode to precisely intervene with personalized phenome-based data. Here, we report the results of a 3-month study on 350 individuals with metabolic syndrome high-risk that we named the Pioneer 350 Wellness project (P350). We examined: (1) longitudinal (two times) phenotypes covering blood lipids, blood glucose, homocysteine (HCY), and vitamin D3(VD3), and (2) polymorphism of genes related to folic acid metabolism. Based on personalized data and questionnaires including demographics, diet and exercise habits information, coaches identified 'actionable possibilities', which combined exercise, diet, and dietary supplements. After a 3-month proactive intervention, two-thirds of the phenotypic markers were significantly improved in the P350 cohort. Specifically, we found that dietary supplements and lifestyle interventions have different effects on phenotypic improvement. For example, dietary supplements can result in a rapid recovery of abnormal HCY and VD3levels, while lifestyle interventions are more suitable for those with high body mass index (BMI), but almost do not help the recovery of HCY. Furthermore, although people who implemented only one of the exercise or diet interventions also benefited, the effect was not as good as the combined exercise and diet interventions. In a subgroup of 226 people, we examined the association between the polymorphism of genes related to folic acid metabolism and the benefits of folate supplementation to restore a normal HCY level. We found people with folic acid metabolism deficiency genes are more likely to benefit from folate supplementation to restore a normal HCY level. Overall, these results suggest: (1) phenome-based data can guide the formulation of more precise and comprehensive interventions, and (2) genetic polymorphism impacts clinical responses to interventions. Notably, we provide a proactive intervention example that is operable in daily life, allowing people with different phenome-based data to design the appropriate intervention protocol including dietary supplements and lifestyle interventions.

As a challenging problem, few-shot class-incremental learning (FSCIL) continually learns a sequence of tasks, confronting the dilemma between slow forgetting of old knowledge and fast adaptation to new knowledge. In this paper, we concentrate on this “slow versus fast” (SvF) dilemma to determine which knowledge components to be updated in a slow fashion or a fast fashion, and thereby balance old-knowledge preservation and new-knowledge adaptation. We propose a multi-grained SvF learning strategy to cope with the SvF dilemma from two different grains: intra-space (within the same feature space) and inter-space (between two different feature spaces). The proposed strategy designs a novel frequency-aware regularization to boost the intra-space SvF capability, and meanwhile develops a new feature space composition operation to enhance the inter-space SvF learning performance. With the multi-grained SvF learning strategy, our method outperforms the state-of-the-art approaches by a large margin.

In large infrastructure projects, the precision and quality of welding large-scale steel mesh sheets are crucial for the safety of the building structure. Traditional manual welding is inefficient and risky; thus, there is an urgent need to replace manual labor with specialized robots equipped with visual recognition systems. However, identifying features like weld points on diverse steel components remains a bottleneck for the application of robotic welding technology. This article proposes a framework for the simultaneous extraction of multitype steel mesh features based on measured point clouds to meet the requirements of a visual-guided robotic welding system. Initially, the Euclidean clustering method is employed to segment the point cloud and obtain all point cloud clusters that belong to the rebar body. Subsequently, the principal component analysis (PCA) and geometric constraints are utilized to merge the point cloud clusters belonging to the same rebar or overlapping rebar. Finally, the feature extraction of rebar’s intersection welding points and lap weld seams is performed using methods such as statistical outlier removal (SOR) and least squares (LSs) cylinder fitting. Additional ablation experiments were conducted to assess the robustness of the framework. To substantiate the feasibility of the framework, the recommended approach was compared with existing point cloud-based and image-based methods. The results indicate that the framework achieves a balance between efficiency and stable robustness while ensuring recognition accuracy.