BACKGROUND In an era leaning toward a personalized alignment of total knee arthroplasty,coronal plane alignment of the knee(CPAK)phenotypes for each population are studied;furthermore,other possible variables affectin...BACKGROUND In an era leaning toward a personalized alignment of total knee arthroplasty,coronal plane alignment of the knee(CPAK)phenotypes for each population are studied;furthermore,other possible variables affecting the alignment,such as ankle joint alignment,should be considered.AIM To determine CPAK distribution in the North African(Egyptian)population with knee osteoarthritis and to assess ankle joint line orientation(AJLO)adaptations across different CPAK types.METHODS A cross-sectional study was conducted on patients with primary knee osteoarthritis and normal ankle joints.Radiographic parameters included the mechanical lateral distal femoral angle,medial proximal tibial angle,and the derived calculations of joint line obliquity(JLO)and arithmetic hip-knee-ankle angle(aHKA).The tibial plafond horizontal angle(TPHA)was used for AJLO assessment,where 0°is neutral(type N),<0°is varus(type A),and>0°is valgus(type B).The nine CPAK types were further divided into 27 subtypes after incorporating the three AJLO types.RESULTS A total of 527 patients(1054 knees)were included for CPAK classification,and 435 patients(870 knees and ankles)for AJLO assessment.The mean age was 57.2±7.8 years,with 79.5%females.Most knees(76.4%)demonstrated varus alignment(mean aHKA was-5.51°±4.84°)and apex distal JLO(55.3%)(mean JLO was 176.43°±4.53°).CPAK types I(44.3%),IV(28.6%),and II(10%)were the most common.Regarding AJLO,70.2%of ankles exhibited varus orientation(mean TPHA was-5.21°±6.45°).The most frequent combined subtypes were CPAK type I-A(33.7%),IV-A(21.5%),and I-N(6.9%).A significant positive correlation was found between the TPHA and aHKA(r=0.40,P<0.001).CONCLUSION In this North African cohort,varus knee alignment with apex distal JLO and varus AJLO predominated.CPAK types I,IV,and II were the most common types,while subtypes I-A,IV-A,and I-N were commonly occurring after incorporating AJLO types;furthermore,the AJLO was significantly correlated to aHKA.展开更多
A polarization-sensitive and flexible photodetector was fabricated through the precise alignment of perovskite nanowires(NWs)using a brush coating technique.The alignment of the NWs was meticulously examined,consideri...A polarization-sensitive and flexible photodetector was fabricated through the precise alignment of perovskite nanowires(NWs)using a brush coating technique.The alignment of the NWs was meticulously examined,considering various chemical properties of the solvent,such as boiling point,viscosity,and surface tension.Notably,when the NWs were brush-coated with toluene dispersion,the NWs were aligned in higher order than those processed from octane dispersion.The degree of alignment was correlated with the photodetector property.Especially,the well-aligned NW photodetector exhibited a two-fold disparity in current response contingent on the polarization direction.Furthermore,even after enduring 500 bending cycles,the device retained 80%of its photodetector performance.This approach underscores the potential of solution-processed flexible photodetectors for advanced optical applications under dynamic operating conditions.展开更多
Nanocrystals have emerged as cutting-edge functional materials benefiting from the increased surface and enhanced coupling of electronic states.High-resolution imaging in transmission electron microscope can provide i...Nanocrystals have emerged as cutting-edge functional materials benefiting from the increased surface and enhanced coupling of electronic states.High-resolution imaging in transmission electron microscope can provide invaluable structural information of crystalline materials,albeit it remains greatly challenging to nanocrystals due to the arduousness of accurate zone axis adjustment.Herein,a homemade software package,called SmartAxis,is developed for rapid yet accurate zone axis alignment of nanocrystals.Incident electron beam tilt is employed as an eccentric goniometer to measure the angular deviation of a crystal to a zone axis,and then serves as a linkage to calculate theαandβtilts of goniometer based on an accurate quantitative relationship.In this way,high-resolution imaging of one identical small Au nanocrystal,as well as electron beam-sensitive MIL-101 metal-organic framework crystals,along multiple zone axes,was performed successfully by using this accurate,time-and electron dose-saving zone axis alignment software package.展开更多
With the rapid expansion of social media,analyzing emotions and their causes in texts has gained significant importance.Emotion-cause pair extraction enables the identification of causal relationships between emotions...With the rapid expansion of social media,analyzing emotions and their causes in texts has gained significant importance.Emotion-cause pair extraction enables the identification of causal relationships between emotions and their triggers within a text,facilitating a deeper understanding of expressed sentiments and their underlying reasons.This comprehension is crucial for making informed strategic decisions in various business and societal contexts.However,recent research approaches employing multi-task learning frameworks for modeling often face challenges such as the inability to simultaneouslymodel extracted features and their interactions,or inconsistencies in label prediction between emotion-cause pair extraction and independent assistant tasks like emotion and cause extraction.To address these issues,this study proposes an emotion-cause pair extraction methodology that incorporates joint feature encoding and task alignment mechanisms.The model consists of two primary components:First,joint feature encoding simultaneously generates features for emotion-cause pairs and clauses,enhancing feature interactions between emotion clauses,cause clauses,and emotion-cause pairs.Second,the task alignment technique is applied to reduce the labeling distance between emotion-cause pair extraction and the two assistant tasks,capturing deep semantic information interactions among tasks.The proposed method is evaluated on a Chinese benchmark corpus using 10-fold cross-validation,assessing key performance metrics such as precision,recall,and F1 score.Experimental results demonstrate that the model achieves an F1 score of 76.05%,surpassing the state-of-the-art by 1.03%.The proposed model exhibits significant improvements in emotion-cause pair extraction(ECPE)and cause extraction(CE)compared to existing methods,validating its effectiveness.This research introduces a novel approach based on joint feature encoding and task alignment mechanisms,contributing to advancements in emotion-cause pair extraction.However,the study’s limitation lies in the data sources,potentially restricting the generalizability of the findings.展开更多
Surface-enhanced Raman spectroscopy(SERS)has evolved from a laboratory technique to a practical tool for ultra-sensitive detection,particularly in the biomedical field,where precise molecular identification is crucial...Surface-enhanced Raman spectroscopy(SERS)has evolved from a laboratory technique to a practical tool for ultra-sensitive detection,particularly in the biomedical field,where precise molecular identification is crucial.Despite significant advancements,a gap remains in the literature,as no comprehensive review systematically addresses the high-precision construction of SERS substrates for ultrasensitive biomedical detection.This review fills that gap by exploring recent progress in fabricating high-precision SERS substrates,emphasizing their role in enabling ultrasensitive bio-medical sensors.We carefully examine the key to these advancements is the precision engineering of substrates,including noble metals,semiconductors,carbon-based materials,and two-dimensional materials,which is essential for achieving the high sensitivity required for ultrasensitive detection.Applications in biomedical diagnostics and molecular analysis are highlighted.Finally,we address the challenges in SERS substrate preparation and outline future directions,focusing on improvement strategies,design concepts,and expanding applications for these advanced materials.展开更多
The band alignment between silicon and high-k dielectrics,which is a key factor in device operation and reliability,still suffers from uncontrolled fluctuations and ambiguous understanding.In this study,by conducting ...The band alignment between silicon and high-k dielectrics,which is a key factor in device operation and reliability,still suffers from uncontrolled fluctuations and ambiguous understanding.In this study,by conducting atomic-level ab initio calculations on realistic Si/SiO_(2)/HfO_(2)stacks,we reveal the physical origin of band alignment fluctuations,i.e.,the oxygen density-dependent interface and surface dipoles,and demonstrate that band offsets can be tuned without introducing other materials.This is instructive for reducing the gate tunneling current,alleviating device-to-device variation,and tuning the threshold voltage.Additionally,this study indicates that significant attention should be focused on model construction in emerging atomistic studies on semiconductor devices.展开更多
Large-aperture optical components are of paramount importance in domains such as integrated circuits,photolithography,aerospace,and inertial confinement fusion.However,measuring their surface profiles relies predomina...Large-aperture optical components are of paramount importance in domains such as integrated circuits,photolithography,aerospace,and inertial confinement fusion.However,measuring their surface profiles relies predominantly on the phase-shifting approach,which involves collecting multiple interferograms and imposes stringent demands on environmental stability.These issues significantly hinder its ability to achieve real-time and dynamic high-precision measurements.Therefore,this study proposes a high-precision large-aperture single-frame interferometric surface profile measurement(LA-SFISPM)method based on deep learning and explores its capability to realize dynamic measurements with high accuracy.The interferogram is matched to the phase by training the data measured using the small aperture.The consistency of the surface features of the small and large apertures is enhanced via contrast learning and feature-distribution alignment.Hence,high-precision phase reconstruction of large-aperture optical components can be achieved without using a phase shifter.The experimental results show that for the tested mirror withΦ=820 mm,the surface profile obtained from LA-SFISPM is subtracted point-by-point from the ground truth,resulting in a maximum single-point error of 4.56 nm.Meanwhile,the peak-to-valley(PV)value is 0.0758λ,and the simple repeatability of root mean square(SR-RMS)value is 0.00025λ,which aligns well with the measured results obtained by ZYGO.In particular,a significant reduction in the measurement time(reduced by a factor of 48)is achieved compared with that of the traditional phase-shifting method.Our proposed method provides an efficient,rapid,and accurate method for obtaining the surface profiles of optical components with different diameters without employing a phase-shifting approach,which is highly desired in large-aperture interferometric measurement systems.展开更多
A high pattern resolution is critical for fabricating roll-to-roll printed electronics(R2RPE)products.For enhanced overlay alignment accuracy,position errors between the printer and the substrate web must be eliminate...A high pattern resolution is critical for fabricating roll-to-roll printed electronics(R2RPE)products.For enhanced overlay alignment accuracy,position errors between the printer and the substrate web must be eliminated,particularly in inkjet printing applications.This paper proposes a novel five-degree-of-freedom(5-DOF)flexure-based alignment stage to adjust the posture of an inkjet printer head.The stage effectively compensates for positioning errors between the actuation mechanism and manipulated objects through a series-parallel combination of compliant substructures.Voice coil motors(VCMs)and linear motors serve as actuators to achieve the required motion.Theoretical models were established using a pseudo-rigid-body model(PRBM)methodology and were validated through finite element analysis(FEA).Finally,an alignment stage prototype was fabricated for an experiment.The prototype test results showed that the developed positioning platform attains 5-DOF motion capabilities with 335.1μm×418.9μm×408.1μm×3.4 mrad×3.29 mrad,with cross-axis coupling errors below 0.11%along y-and z-axes.This paper pro-poses a novel 5-DOF flexure-based alignment stage that can be used for error compensation in R2RPE and effectively improves the interlayer alignment accuracy of multi-layer printing.展开更多
Video classification is an important task in video understanding and plays a pivotal role in intelligent monitoring of information content.Most existing methods do not consider the multimodal nature of the video,and t...Video classification is an important task in video understanding and plays a pivotal role in intelligent monitoring of information content.Most existing methods do not consider the multimodal nature of the video,and the modality fusion approach tends to be too simple,often neglecting modality alignment before fusion.This research introduces a novel dual stream multimodal alignment and fusion network named DMAFNet for classifying short videos.The network uses two unimodal encoder modules to extract features within modalities and exploits a multimodal encoder module to learn interaction between modalities.To solve the modality alignment problem,contrastive learning is introduced between two unimodal encoder modules.Additionally,masked language modeling(MLM)and video text matching(VTM)auxiliary tasks are introduced to improve the interaction between video frames and text modalities through backpropagation of loss functions.Diverse experiments prove the efficiency of DMAFNet in multimodal video classification tasks.Compared with other two mainstream baselines,DMAFNet achieves the best results on the 2022 WeChat Big Data Challenge dataset.展开更多
The elliptic integral method(EIM) is an efficient analytical approach for analyzing large deformations of elastic beams. However, it faces the following challenges.First, the existing EIM can only handle cases with kn...The elliptic integral method(EIM) is an efficient analytical approach for analyzing large deformations of elastic beams. However, it faces the following challenges.First, the existing EIM can only handle cases with known deformation modes. Second,the existing EIM is only applicable to Euler beams, and there is no EIM available for higher-precision Timoshenko and Reissner beams in cases where both force and moment are applied at the end. This paper proposes a general EIM for Reissner beams under arbitrary boundary conditions. On this basis, an analytical equation for determining the sign of the elliptic integral is provided. Based on the equation, we discover a class of elliptic integral piecewise points that are distinct from inflection points. More importantly, we propose an algorithm that automatically calculates the number of inflection points and other piecewise points during the nonlinear solution process, which is crucial for beams with unknown or changing deformation modes.展开更多
Large language models(LLMs)represent significant advancements in artificial intelligence.However,their increasing capabilities come with a serious challenge:misalignment,which refers to the deviation of model behavior...Large language models(LLMs)represent significant advancements in artificial intelligence.However,their increasing capabilities come with a serious challenge:misalignment,which refers to the deviation of model behavior from the designers’intentions and human values.This review aims to synthesize the current understanding of the LLM misalignment issue and provide researchers and practitioners with a comprehensive overview.We define the concept of misalignment and elaborate on its various manifestations,including generating harmful content,factual errors(hallucinations),propagating biases,failing to follow instructions,emerging deceptive behaviors,and emergent misalignment.We explore the multifaceted causes of misalignment,systematically analyzing factors from surface-level technical issues(e.g.,training data,objective function design,model scaling)to deeper fundamental challenges(e.g.,difficulties formalizing values,discrepancies between training signals and real intentions).This review covers existing and emerging techniques for detecting and evaluating the degree of misalignment,such as benchmark tests,red-teaming,and formal safety assessments.Subsequently,we examine strategies to mitigate misalignment,focusing on mainstream alignment techniques such as RLHF,Constitutional AI(CAI),instruction fine-tuning,and novel approaches that address scalability and robustness.In particular,we analyze recent advances in misalignment attack research,including system prompt modifications,supervised fine-tuning,self-supervised representation attacks,and model editing,which challenge the robustness of model alignment.We categorize and analyze the surveyed literature,highlighting major findings,persistent limitations,and current contentious points.Finally,we identify key open questions and propose several promising future research directions,including constructing high-quality alignment datasets,exploring novel alignment methods,coordinating diverse values,and delving into the deep philosophical aspects of alignment.This work underscores the complexity and multidimensionality of LLM misalignment issues,calling for interdisciplinary approaches to reliably align LLMs with human values.展开更多
Bismuth oxyselenide(Bi_(2)O_(2)Se),a novel quasi-two-dimensional charge-carrying semiconductor,is recognized as one of the most promising emerging platforms for next-generation semiconductor devices.Recent advancement...Bismuth oxyselenide(Bi_(2)O_(2)Se),a novel quasi-two-dimensional charge-carrying semiconductor,is recognized as one of the most promising emerging platforms for next-generation semiconductor devices.Recent advancements in the development of diverse Bi_(2)O_(2)Se heterojunctions have unveiled extensive potential applications in both electronics and optoelectronics.However,achieving an in-depth understanding of band alignment and particularly interface dynamics remains a significant challenge.In this study,we conduct a comprehensive experimental investigation into band alignment utilizing high-resolution X-ray photoelectron spectroscopy(HRXPS),while also thoroughly discussing the properties of interface states.Our findings reveal that ultrathin films of Bi_(2)O_(2)Se grown on SrTiO_(3)(with TiO_(2)(001)termination)exhibit Type-I(straddling gap)band alignment characterized by a valence band offset(VBO)of approximately 1.77±0.04 eV and a conduction band offset(CBO)around 0.68±0.04 eV.Notably,when accounting for the influence of interface states,the bands at the interface display a herringbone configuration due to substantial built-in electric fields,which markedly deviate from conventional band alignments.Thus,our results provide valuable insights for advancing high-efficiency electronic and optoelectronic devices,particularly those where charge transfer is highly sensitive to interface states.展开更多
To address the challenge of missing modal information in entity alignment and to mitigate information loss or bias arising frommodal heterogeneity during fusion,while also capturing shared information acrossmodalities...To address the challenge of missing modal information in entity alignment and to mitigate information loss or bias arising frommodal heterogeneity during fusion,while also capturing shared information acrossmodalities,this paper proposes a Multi-modal Pre-synergistic Entity Alignmentmodel based on Cross-modalMutual Information Strategy Optimization(MPSEA).The model first employs independent encoders to process multi-modal features,including text,images,and numerical values.Next,a multi-modal pre-synergistic fusion mechanism integrates graph structural and visual modal features into the textual modality as preparatory information.This pre-fusion strategy enables unified perception of heterogeneous modalities at the model’s initial stage,reducing discrepancies during the fusion process.Finally,using cross-modal deep perception reinforcement learning,the model achieves adaptive multilevel feature fusion between modalities,supporting learningmore effective alignment strategies.Extensive experiments on multiple public datasets show that the MPSEA method achieves gains of up to 7% in Hits@1 and 8.2% in MRR on the FBDB15K dataset,and up to 9.1% in Hits@1 and 7.7% in MRR on the FBYG15K dataset,compared to existing state-of-the-art methods.These results confirm the effectiveness of the proposed model.展开更多
Just a few days after the astronomical spectacle of the“seven planets in alignment”on February 28,2025,the weather in many places underwent a dramatic change.In Shanghai,which was still at the beginning of spring(Ma...Just a few days after the astronomical spectacle of the“seven planets in alignment”on February 28,2025,the weather in many places underwent a dramatic change.In Shanghai,which was still at the beginning of spring(March 1),the temperature suddenly soared to 29℃,the temperature of summer,while Shandong was hit by a sudden heavy snowstorm.There are various opinions and no consensus on this inexplicable weather change.For this reason,based on the principle of the role of planets in the luminescence and heat generation of stars,the author of this article reveals the significant impact of the“Seven planets in alignment”on global climate change,and also points out that the melting of polar glaciers and the approach of the moon to the Earth is another important cause of global climate change.Therefore,countermeasures to save the abnormal changes in global climate are proposed.展开更多
With the intensifying competition in the integrated circuit(IC)industry,the high turnover rate of integrated circuit engineers has become a prominent issue affecting the technological continuity of high-precision,spec...With the intensifying competition in the integrated circuit(IC)industry,the high turnover rate of integrated circuit engineers has become a prominent issue affecting the technological continuity of high-precision,specialized,and innovative enterprises.As a representative of such enterprises,JL Technology has faced challenges to its R&D efficiency due to talent loss in recent years.This study takes this enterprise as a case to explore feasible paths to reduce turnover rates through optimizing training and career development systems.The research designs a method combining learning maps and talent maps,utilizes a competency model to clarify the direction for engineers’skill improvement,implements talent classification management using a nine-grid model,and achieves personalized training through Individual Development Plans(IDPs).Analysis of the enterprise’s historical data reveals that the main reasons for turnover are unclear career development paths and insufficient resources for skill improvement.After pilot implementation,the turnover rate in core departments decreased by 12%,and employee satisfaction with training increased by 24%.The results indicate that matching systematic talent reviews with dynamic learning resources can effectively enhance engineers’sense of belonging.This study provides a set of highly operational management tools for small and medium-sized high-precision,specialized,and innovative technology enterprises,verifies their applicability in such enterprises,and offers replicable experiences for similar enterprises to optimize their talent strategies[1].展开更多
Background The redirected walking(RDW)method for multi-user collaboration requires maintaining the relative position between users in a virtual environment(VE)and physical environment(PE).A chasing game in a VE is a t...Background The redirected walking(RDW)method for multi-user collaboration requires maintaining the relative position between users in a virtual environment(VE)and physical environment(PE).A chasing game in a VE is a typical virtual reality game that entails multi-user collaboration.When a user approaches and interacts with a target user in the VE,the user is expected to approach and interact with the target user in the corresponding PE as well.Existing methods of multi-user RDW mainly focus on obstacle avoidance,which does not account for the relative positional relationship between the users in both VE and PE.Methods To enhance the user experience and facilitate potential interaction,this paper presents a novel dynamic alignment algorithm for multi-user collaborative redirected walking(DA-RDW)in a shared PE where the target user and other users are moving.This algorithm adopts improved artificial potential fields,where the repulsive force is a function of the relative position and velocity of the user with respect to dynamic obstacles.For the best alignment,this algorithm sets the alignment-guidance force in several cases and then converts it into a constrained optimization problem to obtain the optimal direction.Moreover,this algorithm introduces a potential interaction object selection strategy for a dynamically uncertain environment to speed up the subsequent alignment.To balance obstacle avoidance and alignment,this algorithm uses the dynamic weightings of the virtual and physical distances between users and the target to determine the resultant force vector.Results The efficacy of the proposed method was evaluated using a series of simulations and live-user experiments.The experimental results demonstrate that our novel dynamic alignment method for multi-user collaborative redirected walking can reduce the distance error in both VE and PE to improve alignment with fewer collisions.展开更多
This paper deals with the problem of recreating horizontal alignments of existing railway lines.The main objective is to propose a simple method for automatically obtaining optimized recreated alignments located as cl...This paper deals with the problem of recreating horizontal alignments of existing railway lines.The main objective is to propose a simple method for automatically obtaining optimized recreated alignments located as close as possible to an existing one.Based on a previously defined geometric model,two different constrained optimization problems are formulated.The first problem uses only the information provided by a set of points representing the track centerline while the second one also considers additional data about the existing alignment.The proposed methodology consists of a two-stage process in which both problems are solved consecutively using numerical techniques.The main results obtained applying this methodology are presented to show its performance and to prove its practical usefulness:an academic example used to compare with other methods,and a case study of a railway section located in Parga(Spain)in which the geometry of its horizontal alignment is successfully recovered.展开更多
The multiple oligopeptides have been regarded as promising alignment media due to their structural diverseness and tendency for self-assembly in solution.Herein,an assembled amphiphilic peptide alignment medium,i.e.,C...The multiple oligopeptides have been regarded as promising alignment media due to their structural diverseness and tendency for self-assembly in solution.Herein,an assembled amphiphilic peptide alignment medium,i.e.,C15eCONH-Phg-Phg-IIIKK-CONH2 with un-natural amino acids for the determination of anisotropic parameters of NMR is introduced.The amphiphilic peptide can be self-assembled at low concentrations in DMSO and is stable and highly homogeneous.The NMR spectrum collected with the addition of the medium had fewer background signals.The utility of the acquired RDC data is demon-strated to determine relative configuration of three natural products,Helminthosporic acid,Estrone,and a-Santonin.展开更多
With the advancements in parameter-efficient transfer learning techniques,it has become feasible to leverage large pre-trained language models for downstream tasks under low-cost and low-resource conditions.However,ap...With the advancements in parameter-efficient transfer learning techniques,it has become feasible to leverage large pre-trained language models for downstream tasks under low-cost and low-resource conditions.However,applying this technique to multimodal knowledge transfer introduces a significant challenge:ensuring alignment across modalities while minimizing the number of additional parameters required for downstream task adaptation.This paper introduces UniTrans,a framework aimed at facilitating efficient knowledge transfer across multiple modalities.UniTrans leverages Vector-based Cross-modal Random Matrix Adaptation to enable fine-tuning with minimal parameter overhead.To further enhance modality alignment,we introduce two key components:the Multimodal Consistency Alignment Module and the Query-Augmentation Side Network,specifically optimized for scenarios with extremely limited trainable parameters.Extensive evaluations on various cross-modal downstream tasks demonstrate that our approach surpasses state-of-the-art methods while using just 5%of their trainable parameters.Additionally,it achieves superior performance compared to fully fine-tuned models on certain benchmarks.展开更多
Deep learning networks are increasingly exploited in the field of neuronal soma segmentation.However,annotating dataset is also an expensive and time-consuming task.Unsupervised domain adaptation is an effective metho...Deep learning networks are increasingly exploited in the field of neuronal soma segmentation.However,annotating dataset is also an expensive and time-consuming task.Unsupervised domain adaptation is an effective method to mitigate the problem,which is able to learn an adaptive segmentation model by transferring knowledge from a rich-labeled source domain.In this paper,we propose a multi-level distribution alignment-based unsupervised domain adaptation network(MDA-Net)for segmentation of 3D neuronal soma images.Distribution alignment is performed in both feature space and output space.In the feature space,features from different scales are adaptively fused to enhance the feature extraction capability for small target somata and con-strained to be domain invariant by adversarial adaptation strategy.In the output space,local discrepancy maps that can reveal the spatial structures of somata are constructed on the predicted segmentation results.Then thedistribution alignment is performed on the local discrepancies maps across domains to obtain a superior discrepancy map in the target domain,achieving refined segmentation performance of neuronal somata.Additionally,after a period of distribution align-ment procedure,a portion of target samples with high confident pseudo-labels are selected as training data,which assist in learning a more adaptive segmentation network.We verified the superiority of the proposed algorithm by comparing several domain adaptation networks on two 3D mouse brain neuronal somata datasets and one macaque brain neuronal soma dataset.展开更多
基金approved by Institutional Review Board of Faculty of Medicine in Assiut University,No.04-2024-300470.
文摘BACKGROUND In an era leaning toward a personalized alignment of total knee arthroplasty,coronal plane alignment of the knee(CPAK)phenotypes for each population are studied;furthermore,other possible variables affecting the alignment,such as ankle joint alignment,should be considered.AIM To determine CPAK distribution in the North African(Egyptian)population with knee osteoarthritis and to assess ankle joint line orientation(AJLO)adaptations across different CPAK types.METHODS A cross-sectional study was conducted on patients with primary knee osteoarthritis and normal ankle joints.Radiographic parameters included the mechanical lateral distal femoral angle,medial proximal tibial angle,and the derived calculations of joint line obliquity(JLO)and arithmetic hip-knee-ankle angle(aHKA).The tibial plafond horizontal angle(TPHA)was used for AJLO assessment,where 0°is neutral(type N),<0°is varus(type A),and>0°is valgus(type B).The nine CPAK types were further divided into 27 subtypes after incorporating the three AJLO types.RESULTS A total of 527 patients(1054 knees)were included for CPAK classification,and 435 patients(870 knees and ankles)for AJLO assessment.The mean age was 57.2±7.8 years,with 79.5%females.Most knees(76.4%)demonstrated varus alignment(mean aHKA was-5.51°±4.84°)and apex distal JLO(55.3%)(mean JLO was 176.43°±4.53°).CPAK types I(44.3%),IV(28.6%),and II(10%)were the most common.Regarding AJLO,70.2%of ankles exhibited varus orientation(mean TPHA was-5.21°±6.45°).The most frequent combined subtypes were CPAK type I-A(33.7%),IV-A(21.5%),and I-N(6.9%).A significant positive correlation was found between the TPHA and aHKA(r=0.40,P<0.001).CONCLUSION In this North African cohort,varus knee alignment with apex distal JLO and varus AJLO predominated.CPAK types I,IV,and II were the most common types,while subtypes I-A,IV-A,and I-N were commonly occurring after incorporating AJLO types;furthermore,the AJLO was significantly correlated to aHKA.
基金supported by a Commercialization Promotion Agency for R&D Outcomes(COMPA)Grant funded by the Korean Government(Ministry of Science and ICT)(No.RS-2023-00304743)the National Research Foundation of Korea(NRF)Grant funded by the Korean Government(MSIT)(No.2022M3J7A1066428)"Regional Innovation Strategy(RIS)"through the National Research Foundation of Korea(NRF)funded by the Ministry of Education(MOE)(No.2023RIS-008).
文摘A polarization-sensitive and flexible photodetector was fabricated through the precise alignment of perovskite nanowires(NWs)using a brush coating technique.The alignment of the NWs was meticulously examined,considering various chemical properties of the solvent,such as boiling point,viscosity,and surface tension.Notably,when the NWs were brush-coated with toluene dispersion,the NWs were aligned in higher order than those processed from octane dispersion.The degree of alignment was correlated with the photodetector property.Especially,the well-aligned NW photodetector exhibited a two-fold disparity in current response contingent on the polarization direction.Furthermore,even after enduring 500 bending cycles,the device retained 80%of its photodetector performance.This approach underscores the potential of solution-processed flexible photodetectors for advanced optical applications under dynamic operating conditions.
基金supported by the National Key R&D Program of China(No.2021YFA1501002)Thousand Talents Program for Distinguished Young Scholars.X.Li thanks the National Natural Science Foundation of China(No.22309021).
文摘Nanocrystals have emerged as cutting-edge functional materials benefiting from the increased surface and enhanced coupling of electronic states.High-resolution imaging in transmission electron microscope can provide invaluable structural information of crystalline materials,albeit it remains greatly challenging to nanocrystals due to the arduousness of accurate zone axis adjustment.Herein,a homemade software package,called SmartAxis,is developed for rapid yet accurate zone axis alignment of nanocrystals.Incident electron beam tilt is employed as an eccentric goniometer to measure the angular deviation of a crystal to a zone axis,and then serves as a linkage to calculate theαandβtilts of goniometer based on an accurate quantitative relationship.In this way,high-resolution imaging of one identical small Au nanocrystal,as well as electron beam-sensitive MIL-101 metal-organic framework crystals,along multiple zone axes,was performed successfully by using this accurate,time-and electron dose-saving zone axis alignment software package.
文摘With the rapid expansion of social media,analyzing emotions and their causes in texts has gained significant importance.Emotion-cause pair extraction enables the identification of causal relationships between emotions and their triggers within a text,facilitating a deeper understanding of expressed sentiments and their underlying reasons.This comprehension is crucial for making informed strategic decisions in various business and societal contexts.However,recent research approaches employing multi-task learning frameworks for modeling often face challenges such as the inability to simultaneouslymodel extracted features and their interactions,or inconsistencies in label prediction between emotion-cause pair extraction and independent assistant tasks like emotion and cause extraction.To address these issues,this study proposes an emotion-cause pair extraction methodology that incorporates joint feature encoding and task alignment mechanisms.The model consists of two primary components:First,joint feature encoding simultaneously generates features for emotion-cause pairs and clauses,enhancing feature interactions between emotion clauses,cause clauses,and emotion-cause pairs.Second,the task alignment technique is applied to reduce the labeling distance between emotion-cause pair extraction and the two assistant tasks,capturing deep semantic information interactions among tasks.The proposed method is evaluated on a Chinese benchmark corpus using 10-fold cross-validation,assessing key performance metrics such as precision,recall,and F1 score.Experimental results demonstrate that the model achieves an F1 score of 76.05%,surpassing the state-of-the-art by 1.03%.The proposed model exhibits significant improvements in emotion-cause pair extraction(ECPE)and cause extraction(CE)compared to existing methods,validating its effectiveness.This research introduces a novel approach based on joint feature encoding and task alignment mechanisms,contributing to advancements in emotion-cause pair extraction.However,the study’s limitation lies in the data sources,potentially restricting the generalizability of the findings.
基金supported by the projects funded by the Education Department of Shaanxi Provincial Government(NO.23JP116)the Natural Science Fund of Shaanxi Province(NO.2024JC-YBMS-396)+3 种基金the National Natural Science Foundation of China(NO.52171191,52371198,U22A20137)the Constructing National Independent Innovation Demonstration Zones(XM2024XTGXQ05)Shenzhen Science and Technology Innovation Program(JCYJ20220818102215033,GJHZ20210705142542015,JCYJ20220530160811027)Guangdong HUST Industrial Technology Research Institute,Guangdong Provincial Key Laboratory of Manufacturing Equipment Digitization(2023B1212060012).
文摘Surface-enhanced Raman spectroscopy(SERS)has evolved from a laboratory technique to a practical tool for ultra-sensitive detection,particularly in the biomedical field,where precise molecular identification is crucial.Despite significant advancements,a gap remains in the literature,as no comprehensive review systematically addresses the high-precision construction of SERS substrates for ultrasensitive biomedical detection.This review fills that gap by exploring recent progress in fabricating high-precision SERS substrates,emphasizing their role in enabling ultrasensitive bio-medical sensors.We carefully examine the key to these advancements is the precision engineering of substrates,including noble metals,semiconductors,carbon-based materials,and two-dimensional materials,which is essential for achieving the high sensitivity required for ultrasensitive detection.Applications in biomedical diagnostics and molecular analysis are highlighted.Finally,we address the challenges in SERS substrate preparation and outline future directions,focusing on improvement strategies,design concepts,and expanding applications for these advanced materials.
基金supported by the National Natural Science Foundation of China(Grant Nos.62174155,12334005,and T2293702)the CAS Project for Young Scientists in Basic Research(Grant No.YSBR-056)the MIND Project(Grant No.MINDKT202403)。
文摘The band alignment between silicon and high-k dielectrics,which is a key factor in device operation and reliability,still suffers from uncontrolled fluctuations and ambiguous understanding.In this study,by conducting atomic-level ab initio calculations on realistic Si/SiO_(2)/HfO_(2)stacks,we reveal the physical origin of band alignment fluctuations,i.e.,the oxygen density-dependent interface and surface dipoles,and demonstrate that band offsets can be tuned without introducing other materials.This is instructive for reducing the gate tunneling current,alleviating device-to-device variation,and tuning the threshold voltage.Additionally,this study indicates that significant attention should be focused on model construction in emerging atomistic studies on semiconductor devices.
基金funded by the National Natural Science Foundation of China Instrumentation Program(52327806)Youth Fund of the National Nature Foundation of China(62405020)China Postdoctoral Science Foundation(2024M764131).
文摘Large-aperture optical components are of paramount importance in domains such as integrated circuits,photolithography,aerospace,and inertial confinement fusion.However,measuring their surface profiles relies predominantly on the phase-shifting approach,which involves collecting multiple interferograms and imposes stringent demands on environmental stability.These issues significantly hinder its ability to achieve real-time and dynamic high-precision measurements.Therefore,this study proposes a high-precision large-aperture single-frame interferometric surface profile measurement(LA-SFISPM)method based on deep learning and explores its capability to realize dynamic measurements with high accuracy.The interferogram is matched to the phase by training the data measured using the small aperture.The consistency of the surface features of the small and large apertures is enhanced via contrast learning and feature-distribution alignment.Hence,high-precision phase reconstruction of large-aperture optical components can be achieved without using a phase shifter.The experimental results show that for the tested mirror withΦ=820 mm,the surface profile obtained from LA-SFISPM is subtracted point-by-point from the ground truth,resulting in a maximum single-point error of 4.56 nm.Meanwhile,the peak-to-valley(PV)value is 0.0758λ,and the simple repeatability of root mean square(SR-RMS)value is 0.00025λ,which aligns well with the measured results obtained by ZYGO.In particular,a significant reduction in the measurement time(reduced by a factor of 48)is achieved compared with that of the traditional phase-shifting method.Our proposed method provides an efficient,rapid,and accurate method for obtaining the surface profiles of optical components with different diameters without employing a phase-shifting approach,which is highly desired in large-aperture interferometric measurement systems.
基金Supported by Natural Science Research Project of Anhui Educational Committee(Grant No.2024AH040010).
文摘A high pattern resolution is critical for fabricating roll-to-roll printed electronics(R2RPE)products.For enhanced overlay alignment accuracy,position errors between the printer and the substrate web must be eliminated,particularly in inkjet printing applications.This paper proposes a novel five-degree-of-freedom(5-DOF)flexure-based alignment stage to adjust the posture of an inkjet printer head.The stage effectively compensates for positioning errors between the actuation mechanism and manipulated objects through a series-parallel combination of compliant substructures.Voice coil motors(VCMs)and linear motors serve as actuators to achieve the required motion.Theoretical models were established using a pseudo-rigid-body model(PRBM)methodology and were validated through finite element analysis(FEA).Finally,an alignment stage prototype was fabricated for an experiment.The prototype test results showed that the developed positioning platform attains 5-DOF motion capabilities with 335.1μm×418.9μm×408.1μm×3.4 mrad×3.29 mrad,with cross-axis coupling errors below 0.11%along y-and z-axes.This paper pro-poses a novel 5-DOF flexure-based alignment stage that can be used for error compensation in R2RPE and effectively improves the interlayer alignment accuracy of multi-layer printing.
基金Fundamental Research Funds for the Central Universities,China(No.2232021A-10)National Natural Science Foundation of China(No.61903078)+1 种基金Shanghai Sailing Program,China(No.22YF1401300)Natural Science Foundation of Shanghai,China(No.20ZR1400400)。
文摘Video classification is an important task in video understanding and plays a pivotal role in intelligent monitoring of information content.Most existing methods do not consider the multimodal nature of the video,and the modality fusion approach tends to be too simple,often neglecting modality alignment before fusion.This research introduces a novel dual stream multimodal alignment and fusion network named DMAFNet for classifying short videos.The network uses two unimodal encoder modules to extract features within modalities and exploits a multimodal encoder module to learn interaction between modalities.To solve the modality alignment problem,contrastive learning is introduced between two unimodal encoder modules.Additionally,masked language modeling(MLM)and video text matching(VTM)auxiliary tasks are introduced to improve the interaction between video frames and text modalities through backpropagation of loss functions.Diverse experiments prove the efficiency of DMAFNet in multimodal video classification tasks.Compared with other two mainstream baselines,DMAFNet achieves the best results on the 2022 WeChat Big Data Challenge dataset.
基金supported by the National Natural Science Foundation of China (Nos. 12172388 and 12472400)the Guangdong Basic and Applied Basic Research Foundation of China(No. 2025A1515011975)the Scientific Research Project of Guangdong Polytechnic Normal University of China (No. 2023SDKYA010)
文摘The elliptic integral method(EIM) is an efficient analytical approach for analyzing large deformations of elastic beams. However, it faces the following challenges.First, the existing EIM can only handle cases with known deformation modes. Second,the existing EIM is only applicable to Euler beams, and there is no EIM available for higher-precision Timoshenko and Reissner beams in cases where both force and moment are applied at the end. This paper proposes a general EIM for Reissner beams under arbitrary boundary conditions. On this basis, an analytical equation for determining the sign of the elliptic integral is provided. Based on the equation, we discover a class of elliptic integral piecewise points that are distinct from inflection points. More importantly, we propose an algorithm that automatically calculates the number of inflection points and other piecewise points during the nonlinear solution process, which is crucial for beams with unknown or changing deformation modes.
基金supported by National Natural Science Foundation of China(62462019,62172350)Guangdong Basic and Applied Basic Research Foundation(2023A1515012846)+6 种基金Guangxi Science and Technology Major Program(AA24263010)The Key Research and Development Program of Guangxi(AB24010085)Key Laboratory of Equipment Data Security and Guarantee Technology,Ministry of Education(GDZB2024060500)2024 Higher Education Scientific Research Planning Project(No.24NL0419)Nantong Science and Technology Project(No.JC2023070)the Open Fund of Advanced Cryptography and System Security Key Laboratory of Sichuan Province(GrantNo.SKLACSS-202407)sponsored by the Cultivation of Young andMiddle-aged Academic Leaders in the“Qing Lan Project”of Jiangsu Province and the 2025 Outstanding Teaching Team in the“Qing Lan Project”of Jiangsu Province.
文摘Large language models(LLMs)represent significant advancements in artificial intelligence.However,their increasing capabilities come with a serious challenge:misalignment,which refers to the deviation of model behavior from the designers’intentions and human values.This review aims to synthesize the current understanding of the LLM misalignment issue and provide researchers and practitioners with a comprehensive overview.We define the concept of misalignment and elaborate on its various manifestations,including generating harmful content,factual errors(hallucinations),propagating biases,failing to follow instructions,emerging deceptive behaviors,and emergent misalignment.We explore the multifaceted causes of misalignment,systematically analyzing factors from surface-level technical issues(e.g.,training data,objective function design,model scaling)to deeper fundamental challenges(e.g.,difficulties formalizing values,discrepancies between training signals and real intentions).This review covers existing and emerging techniques for detecting and evaluating the degree of misalignment,such as benchmark tests,red-teaming,and formal safety assessments.Subsequently,we examine strategies to mitigate misalignment,focusing on mainstream alignment techniques such as RLHF,Constitutional AI(CAI),instruction fine-tuning,and novel approaches that address scalability and robustness.In particular,we analyze recent advances in misalignment attack research,including system prompt modifications,supervised fine-tuning,self-supervised representation attacks,and model editing,which challenge the robustness of model alignment.We categorize and analyze the surveyed literature,highlighting major findings,persistent limitations,and current contentious points.Finally,we identify key open questions and propose several promising future research directions,including constructing high-quality alignment datasets,exploring novel alignment methods,coordinating diverse values,and delving into the deep philosophical aspects of alignment.This work underscores the complexity and multidimensionality of LLM misalignment issues,calling for interdisciplinary approaches to reliably align LLMs with human values.
基金supported by the National Natural Science Foundation of China(Nos.52072059,12304078,12274061 and 11774044)the Natural Science Foundation of Sichuan Province(No.2024NSFSC1384).
文摘Bismuth oxyselenide(Bi_(2)O_(2)Se),a novel quasi-two-dimensional charge-carrying semiconductor,is recognized as one of the most promising emerging platforms for next-generation semiconductor devices.Recent advancements in the development of diverse Bi_(2)O_(2)Se heterojunctions have unveiled extensive potential applications in both electronics and optoelectronics.However,achieving an in-depth understanding of band alignment and particularly interface dynamics remains a significant challenge.In this study,we conduct a comprehensive experimental investigation into band alignment utilizing high-resolution X-ray photoelectron spectroscopy(HRXPS),while also thoroughly discussing the properties of interface states.Our findings reveal that ultrathin films of Bi_(2)O_(2)Se grown on SrTiO_(3)(with TiO_(2)(001)termination)exhibit Type-I(straddling gap)band alignment characterized by a valence band offset(VBO)of approximately 1.77±0.04 eV and a conduction band offset(CBO)around 0.68±0.04 eV.Notably,when accounting for the influence of interface states,the bands at the interface display a herringbone configuration due to substantial built-in electric fields,which markedly deviate from conventional band alignments.Thus,our results provide valuable insights for advancing high-efficiency electronic and optoelectronic devices,particularly those where charge transfer is highly sensitive to interface states.
基金partially supported by the National Natural Science Foundation of China under Grants 62471493 and 62402257(for conceptualization and investigation)partially supported by the Natural Science Foundation of Shandong Province,China under Grants ZR2023LZH017,ZR2024MF066,and 2023QF025(for formal analysis and validation)+1 种基金partially supported by the Open Foundation of Key Laboratory of Computing Power Network and Information Security,Ministry of Education,Qilu University of Technology(Shandong Academy of Sciences)under Grant 2023ZD010(for methodology and model design)partially supported by the Russian Science Foundation(RSF)Project under Grant 22-71-10095-P(for validation and results verification).
文摘To address the challenge of missing modal information in entity alignment and to mitigate information loss or bias arising frommodal heterogeneity during fusion,while also capturing shared information acrossmodalities,this paper proposes a Multi-modal Pre-synergistic Entity Alignmentmodel based on Cross-modalMutual Information Strategy Optimization(MPSEA).The model first employs independent encoders to process multi-modal features,including text,images,and numerical values.Next,a multi-modal pre-synergistic fusion mechanism integrates graph structural and visual modal features into the textual modality as preparatory information.This pre-fusion strategy enables unified perception of heterogeneous modalities at the model’s initial stage,reducing discrepancies during the fusion process.Finally,using cross-modal deep perception reinforcement learning,the model achieves adaptive multilevel feature fusion between modalities,supporting learningmore effective alignment strategies.Extensive experiments on multiple public datasets show that the MPSEA method achieves gains of up to 7% in Hits@1 and 8.2% in MRR on the FBDB15K dataset,and up to 9.1% in Hits@1 and 7.7% in MRR on the FBYG15K dataset,compared to existing state-of-the-art methods.These results confirm the effectiveness of the proposed model.
文摘Just a few days after the astronomical spectacle of the“seven planets in alignment”on February 28,2025,the weather in many places underwent a dramatic change.In Shanghai,which was still at the beginning of spring(March 1),the temperature suddenly soared to 29℃,the temperature of summer,while Shandong was hit by a sudden heavy snowstorm.There are various opinions and no consensus on this inexplicable weather change.For this reason,based on the principle of the role of planets in the luminescence and heat generation of stars,the author of this article reveals the significant impact of the“Seven planets in alignment”on global climate change,and also points out that the melting of polar glaciers and the approach of the moon to the Earth is another important cause of global climate change.Therefore,countermeasures to save the abnormal changes in global climate are proposed.
文摘With the intensifying competition in the integrated circuit(IC)industry,the high turnover rate of integrated circuit engineers has become a prominent issue affecting the technological continuity of high-precision,specialized,and innovative enterprises.As a representative of such enterprises,JL Technology has faced challenges to its R&D efficiency due to talent loss in recent years.This study takes this enterprise as a case to explore feasible paths to reduce turnover rates through optimizing training and career development systems.The research designs a method combining learning maps and talent maps,utilizes a competency model to clarify the direction for engineers’skill improvement,implements talent classification management using a nine-grid model,and achieves personalized training through Individual Development Plans(IDPs).Analysis of the enterprise’s historical data reveals that the main reasons for turnover are unclear career development paths and insufficient resources for skill improvement.After pilot implementation,the turnover rate in core departments decreased by 12%,and employee satisfaction with training increased by 24%.The results indicate that matching systematic talent reviews with dynamic learning resources can effectively enhance engineers’sense of belonging.This study provides a set of highly operational management tools for small and medium-sized high-precision,specialized,and innovative technology enterprises,verifies their applicability in such enterprises,and offers replicable experiences for similar enterprises to optimize their talent strategies[1].
基金Supported by STI 2030 Major Projects of China(2021ZD0200400).
文摘Background The redirected walking(RDW)method for multi-user collaboration requires maintaining the relative position between users in a virtual environment(VE)and physical environment(PE).A chasing game in a VE is a typical virtual reality game that entails multi-user collaboration.When a user approaches and interacts with a target user in the VE,the user is expected to approach and interact with the target user in the corresponding PE as well.Existing methods of multi-user RDW mainly focus on obstacle avoidance,which does not account for the relative positional relationship between the users in both VE and PE.Methods To enhance the user experience and facilitate potential interaction,this paper presents a novel dynamic alignment algorithm for multi-user collaborative redirected walking(DA-RDW)in a shared PE where the target user and other users are moving.This algorithm adopts improved artificial potential fields,where the repulsive force is a function of the relative position and velocity of the user with respect to dynamic obstacles.For the best alignment,this algorithm sets the alignment-guidance force in several cases and then converts it into a constrained optimization problem to obtain the optimal direction.Moreover,this algorithm introduces a potential interaction object selection strategy for a dynamically uncertain environment to speed up the subsequent alignment.To balance obstacle avoidance and alignment,this algorithm uses the dynamic weightings of the virtual and physical distances between users and the target to determine the resultant force vector.Results The efficacy of the proposed method was evaluated using a series of simulations and live-user experiments.The experimental results demonstrate that our novel dynamic alignment method for multi-user collaborative redirected walking can reduce the distance error in both VE and PE to improve alignment with fewer collisions.
基金founded by project TED2021129324B-I00 of the Ministerio de Ciencia e Innovación(Spain)and NextGenerationEU(European Union)the Collaboration Agreement between Consellería de Educación,Formación Profesional e Universidades(Xunta de Galicia,Spain)and Universidade de Santiago de Compostela(Spain)which regulates the Specialization Campus Campus Terra under Grant number 2022-PU014support given by Xunta de Galicia(Spain)by means of the research projects 2023 GPC GI-2084 ED431B2023/17 and GRC GI-1563-ED431C 2021/15,respectively.
文摘This paper deals with the problem of recreating horizontal alignments of existing railway lines.The main objective is to propose a simple method for automatically obtaining optimized recreated alignments located as close as possible to an existing one.Based on a previously defined geometric model,two different constrained optimization problems are formulated.The first problem uses only the information provided by a set of points representing the track centerline while the second one also considers additional data about the existing alignment.The proposed methodology consists of a two-stage process in which both problems are solved consecutively using numerical techniques.The main results obtained applying this methodology are presented to show its performance and to prove its practical usefulness:an academic example used to compare with other methods,and a case study of a railway section located in Parga(Spain)in which the geometry of its horizontal alignment is successfully recovered.
基金supported by the National Natural Science Foundation of China(21874158)the Science and Technology Major Program of Gansu Province of China(22ZD6FA006 and 23ZDFA015)+1 种基金We are also grateful for the financial support from the Science and Technology Program of Henan Province(232102311180)the foundation for the University Young Key Teacher of Henan Province(2024GGJS116).
文摘The multiple oligopeptides have been regarded as promising alignment media due to their structural diverseness and tendency for self-assembly in solution.Herein,an assembled amphiphilic peptide alignment medium,i.e.,C15eCONH-Phg-Phg-IIIKK-CONH2 with un-natural amino acids for the determination of anisotropic parameters of NMR is introduced.The amphiphilic peptide can be self-assembled at low concentrations in DMSO and is stable and highly homogeneous.The NMR spectrum collected with the addition of the medium had fewer background signals.The utility of the acquired RDC data is demon-strated to determine relative configuration of three natural products,Helminthosporic acid,Estrone,and a-Santonin.
文摘With the advancements in parameter-efficient transfer learning techniques,it has become feasible to leverage large pre-trained language models for downstream tasks under low-cost and low-resource conditions.However,applying this technique to multimodal knowledge transfer introduces a significant challenge:ensuring alignment across modalities while minimizing the number of additional parameters required for downstream task adaptation.This paper introduces UniTrans,a framework aimed at facilitating efficient knowledge transfer across multiple modalities.UniTrans leverages Vector-based Cross-modal Random Matrix Adaptation to enable fine-tuning with minimal parameter overhead.To further enhance modality alignment,we introduce two key components:the Multimodal Consistency Alignment Module and the Query-Augmentation Side Network,specifically optimized for scenarios with extremely limited trainable parameters.Extensive evaluations on various cross-modal downstream tasks demonstrate that our approach surpasses state-of-the-art methods while using just 5%of their trainable parameters.Additionally,it achieves superior performance compared to fully fine-tuned models on certain benchmarks.
基金supported by the Fund of Key Laboratory of Biomedical Engineering of Hainan Province(No.BME20240001)the STI2030-Major Projects(No.2021ZD0200104)the National Natural Science Foundations of China under Grant 61771437.
文摘Deep learning networks are increasingly exploited in the field of neuronal soma segmentation.However,annotating dataset is also an expensive and time-consuming task.Unsupervised domain adaptation is an effective method to mitigate the problem,which is able to learn an adaptive segmentation model by transferring knowledge from a rich-labeled source domain.In this paper,we propose a multi-level distribution alignment-based unsupervised domain adaptation network(MDA-Net)for segmentation of 3D neuronal soma images.Distribution alignment is performed in both feature space and output space.In the feature space,features from different scales are adaptively fused to enhance the feature extraction capability for small target somata and con-strained to be domain invariant by adversarial adaptation strategy.In the output space,local discrepancy maps that can reveal the spatial structures of somata are constructed on the predicted segmentation results.Then thedistribution alignment is performed on the local discrepancies maps across domains to obtain a superior discrepancy map in the target domain,achieving refined segmentation performance of neuronal somata.Additionally,after a period of distribution align-ment procedure,a portion of target samples with high confident pseudo-labels are selected as training data,which assist in learning a more adaptive segmentation network.We verified the superiority of the proposed algorithm by comparing several domain adaptation networks on two 3D mouse brain neuronal somata datasets and one macaque brain neuronal soma dataset.
