Deep neural networks remain susceptible to adversarial examples,where the goal of an adversarial attack is to introduce small perturbations to the original examples in order to confuse the model without being easily d...Deep neural networks remain susceptible to adversarial examples,where the goal of an adversarial attack is to introduce small perturbations to the original examples in order to confuse the model without being easily detected.Although many adversarial attack methods produce adversarial examples that have achieved great results in the whitebox setting,they exhibit low transferability in the black-box setting.In order to improve the transferability along the baseline of the gradient-based attack technique,we present a novel Stochastic Gradient Accumulation Momentum Iterative Attack(SAMI-FGSM)in this study.In particular,during each iteration,the gradient information is calculated using a normal sampling approach that randomly samples around the sample points,with the highest probability of capturing adversarial features.Meanwhile,the accumulated information of the sampled gradient from the previous iteration is further considered to modify the current updated gradient,and the original gradient attack direction is changed to ensure that the updated gradient direction is more stable.Comprehensive experiments conducted on the ImageNet dataset show that our method outperforms existing state-of-the-art gradient-based attack techniques,achieving an average improvement of 10.2%in transferability.展开更多
Lithography is a Key enabling technique in modern micro/nano scale technology.Achieving the optimal trade-off between resolution,throughput,and cost remains a central focus in the ongoing development.However,current l...Lithography is a Key enabling technique in modern micro/nano scale technology.Achieving the optimal trade-off between resolution,throughput,and cost remains a central focus in the ongoing development.However,current lithographic techniques such as direct-write,projection,and extreme ultraviolet lithography achieve higher resolution at the expense of increased complexity in optical systems or the use of shorter-wavelength light sources,thus raising the overall cost of production.Here,we present a cost-effective and wafer-level perfect conformal contact lithography at the diffraction limit.By leveraging a transferable photoresist,the technique ensures optimal contact between the mask and photoresist with zero-gap,facilitating the transfer of patterns at the diffraction limit while maintaining high fidelity and uniformity across large wafers.This technique applies to a wide range of complex surfaces,including non-conductive glass surfaces,flexible substrates,and curved surfaces.The proposed technique expands the potential of contact photolithography for novel device architectures and practic al manufacturing processes.展开更多
Whether the multi-biological toxicity from lead exposure could be transferred to progeny has not been clarified. In the present study, we explored the Caenorhabditis elegans to analyze the multiple toxicities from lea...Whether the multi-biological toxicity from lead exposure could be transferred to progeny has not been clarified. In the present study, we explored the Caenorhabditis elegans to analyze the multiple toxicities from lead exposure and their possibly transferable properties. The lead exposure could cause series of severe multi-biological defects with a concentration-dependent manner by affecting the endpoints of life span, development, reproduction and locomotion behaviors in nematodes. Moreover, most of these toxicities could be transferred to progeny from lead exposed animals and some of the defects in progeny appeared even more severe than in their parents, such as the body sizes and mean life spans. We summarized the defects caused by lead exposure into three groups according to their transferable properties or rescue patterns. That is, the defects caused by lead exposure could be largely, or partially, or became even more severe in progeny animals. Therefore, our results suggest that lead exposure can cause severely multi-biological defects, and most of these multiple toxicities can be considered as transferable for exposed animals in C. elegans.展开更多
Based on a new idea for research on cycling of marine biogenic elements, this study showed that only the leachable form phosphorus in natural grain sizes marine sediments constitutes the transferable phosphorous in th...Based on a new idea for research on cycling of marine biogenic elements, this study showed that only the leachable form phosphorus in natural grain sizes marine sediments constitutes the transferable phosphorous in the sediments. The transferable phosphorus content in the natural grain sizes surface sediments in the Huanghe River estuary adjacent waters ranges from 58.5-69.8 μg/g, accounting for only 9.1%-11.0% of the total phosphorus content, whereas the leachable form (“transferable") phosphorus content in the sediment after it was totally ground into powder was found to be 454.8-529.2 μg/g, accounting for 73.4%-89.1% of the total phosphorus. Analysis of the correlation between the biomass of benthos and the leachable form (“transferable") phosphorus showed that most of the leachable form (“transferable") phosphorus in the totally ground sediment did not participate in the marine biogeochemical cycling. Furthermore, a synchronous survey on benthos showed that the biomass of meio and macro benthos exhibited good positive correlation with the leachable form of phosphorus in the natural grain sizes sediment, but poorer correlation with the leachable form (“transferable") phosphorus in the totally ground sediment, indicating that transferable phosphorus in marine sediment is the leachable form of phosphorus in the natural grain sizes sediments, and is not the previously known leachable form (“transferable") phosphorus obtained from the totally ground sediment.展开更多
ZnO thin films were deposited on graphite substrates by ultrasonic spray pyrolysis method with Zn（CH3COO）2·2H2O aqueous solution as precursor. The crystalline structure, morphology, and optical properties of th...ZnO thin films were deposited on graphite substrates by ultrasonic spray pyrolysis method with Zn（CH3COO）2·2H2O aqueous solution as precursor. The crystalline structure, morphology, and optical properties of the as-grown ZnO films were investigated systematically as a function of deposition temperature and growth time. Near-band edge ultraviolet （UV） emission was observed in room temperature photoluminescence spectra for the optimized samples, yet the usually observed defect related deep level emissions were nearly undetectable, indicating that high optical quality ZnO thin films could be achieved via this ultrasonic spray pyrolysis method. Considering the features of transferable and low thermal resistance of the graphite substrates, the achievement will be of special interest for the development of high-power semiconductor devices with sufficient vower durability.展开更多
In multi-agent reinforcement learning(MARL),the behaviors of each agent can influence the learning of others,and the agents have to search in an exponentially enlarged joint-action space.Hence,it is challenging for th...In multi-agent reinforcement learning(MARL),the behaviors of each agent can influence the learning of others,and the agents have to search in an exponentially enlarged joint-action space.Hence,it is challenging for the multi-agent teams to explore in the environment.Agents may achieve suboptimal policies and fail to solve some complex tasks.To improve the exploring efficiency as well as the performance of MARL tasks,in this paper,we propose a new approach by transferring the knowledge across tasks.Differently from the traditional MARL algorithms,we first assume that the reward functions can be computed by linear combinations of a shared feature function and a set of taskspecific weights.Then,we define a set of basic MARL tasks in the source domain and pre-train them as the basic knowledge for further use.Finally,once the weights for target tasks are available,it will be easier to get a well-performed policy to explore in the target domain.Hence,the learning process of agents for target tasks is speeded up by taking full use of the basic knowledge that was learned previously.We evaluate the proposed algorithm on two challenging MARL tasks:cooperative boxpushing and non-monotonic predator-prey.The experiment results have demonstrated the improved performance compared with state-of-the-art MARL algorithms.展开更多
Effective fault diagnosis of planetary gearboxes is critical for ensuring the safety and dependability of mechanical drive systems.Nevertheless,variable conditions and inadequate fault data bring huge challenges to it...Effective fault diagnosis of planetary gearboxes is critical for ensuring the safety and dependability of mechanical drive systems.Nevertheless,variable conditions and inadequate fault data bring huge challenges to its practical fault diagnosis.Taking this into account,this study presents a new intelligent fault diagnosis(IFD)approach for planetary gearbox using a transferable deep Q network(TDQN)that merges deep reinforcement learning(DRL)and transfer learning(TL).First,a DRL environment simulation is designed by a predefined classification Markov decision process.Then,leveraging varied-size convolutions and residual learning,a multiscale residual convolutional neural network agent for TDQN is created to automatically learn meaningful features directly from vibration signals while avoiding model degradation.Next,a large source dataset is obtained from complex conditions,and this agent learns an IFD policy via autonomous interaction with the data environment.Finally,a parameter-based TL strategy is adopted to retrain the model on target datasets with variable conditions and small training data,which is conducted by fine-tuning the model parameters gained from the source task to accomplish target tasks.The results show that this TDQN outperforms not only state-of-the-art methods in a source task with an accuracy of 98.53%but also in two target tasks with 99.63%and 98.37%,respectively.展开更多
In the coal-to-ethylene glycol(CTEG)process,precisely estimating quality variables is crucial for process monitoring,optimization,and control.A significant challenge in this regard is relying on offline laboratory ana...In the coal-to-ethylene glycol(CTEG)process,precisely estimating quality variables is crucial for process monitoring,optimization,and control.A significant challenge in this regard is relying on offline laboratory analysis to obtain these variables,which often incurs substantial monetary costs and significant time delays.The resulting few-shot learning scenarios present a hurdle to the efficient development of predictive models.To address this issue,our study introduces the transferable adversarial slow feature extraction network(TASF-Net),an innovative approach designed specifically for few-shot quality prediction in the CTEG process.TASF-Net uniquely integrates the slowness principle with a deep Bayesian framework,effectively capturing the nonlinear and inertial characteristics of the CTEG process.Additionally,the model employs a variable attention mechanism to identify quality-related input variables adaptively at each time step.A key strength of TASF-Net lies in its ability to navigate the complex measurement noise,outliers,and system interference typical in CTEG data.Adversarial learning strategy using a min-max game is adopted to improve its robustness and ability to model irregular industrial data accurately and significantly.Furthermore,an incremental refining transfer learning framework is designed to further improve few-shot prediction performance achieved by transferring knowledge from the pretrained model on the source domain to the target domain.The effectiveness and superiority of TASF-Net have been empirically validated using a real-world CTEG dataset.Compared with some state-of-the-art methods,TASF-Net demonstrates exceptional capability in addressing the intricate challenges for few-shot quality prediction in the CTEG process.展开更多
In this paper, a new cash scheme is proposed for electronic payment system, in which the cash can be transferred several times. When this kind of cash is used, the fraud such as double spending can be found out but th...In this paper, a new cash scheme is proposed for electronic payment system, in which the cash can be transferred several times. When this kind of cash is used, the fraud such as double spending can be found out but the bank and the trusted party needs not be involved online in each transaction. This cash system is anonymous in normal transactions. But if a fraud happens, the trusted party can withdraw the anonymity to find out the cheater. The new cash scheme is transferable, anonymous, off-line and efficient.展开更多
With the development of information technology,malware threats to the industrial system have become an emergent issue,since various industrial infrastructures have been deeply integrated into our modern works and live...With the development of information technology,malware threats to the industrial system have become an emergent issue,since various industrial infrastructures have been deeply integrated into our modern works and lives.To identify and classify new malware variants,different types of deep learning models have been widely explored recently.Generally,sufficient data is usually required to achieve a well-trained deep learning classifier with satisfactory generalization ability.However,in current practical applications,an ample supply of data is absent in most specific industrial malware detection scenarios.Transfer learning as an effective approach can be used to alleviate the influence of the small sample size problem.In addition,it can also reuse the knowledge from pretrained models,which is beneficial to the real-time requirement in industrial malware detection.In this paper,we investigate the transferable features learned by a 1D-convolutional network and evaluate our proposed methods on 6 transfer learning tasks.The experiment results show that 1D-convolutional architecture is effective to learn transferable features for malware classification,and indicate that transferring the first 2 layers of our proposed 1D-convolutional network is the most efficient way to reuse the learned features.展开更多
In recent years,deep learning(DL)models have achieved signifcant progress in many domains,such as autonomous driving,facial recognition,and speech recognition.However,the vulnerability of deep learning models to adver...In recent years,deep learning(DL)models have achieved signifcant progress in many domains,such as autonomous driving,facial recognition,and speech recognition.However,the vulnerability of deep learning models to adversarial attacks has raised serious concerns in the community because of their insufcient robustness and generalization.Also,transferable attacks have become a prominent method for black-box attacks.In this work,we explore the potential factors that impact adversarial examples(AEs)transferability in DL-based speech recognition.We also discuss the vulnerability of diferent DL systems and the irregular nature of decision boundaries.Our results show a remarkable diference in the transferability of AEs between speech and images,with the data relevance being low in images but opposite in speech recognition.Motivated by dropout-based ensemble approaches,we propose random gradient ensembles and dynamic gradient-weighted ensembles,and we evaluate the impact of ensembles on the transferability of AEs.The results show that the AEs created by both approaches are valid for transfer to the black box API.展开更多
This paper focuses on an important type of black-box attacks,i.e.,transfer-based adversarial attacks,where the adversary generates adversarial examples using a substitute(source)model and utilizes them to attack an un...This paper focuses on an important type of black-box attacks,i.e.,transfer-based adversarial attacks,where the adversary generates adversarial examples using a substitute(source)model and utilizes them to attack an unseen target model,without knowing its information.Existing methods tend to give unsatisfactory adversarial transferability when the source and target models are from different types of DNN architectures(e.g.,ResNet-18 and Swin Transformer).In this paper,we observe that the above phenomenon is induced by the output inconsistency problem.To alleviate this problem while effectively utilizing the existing DNN models,we propose a common knowledge learning(CKL)framework to learn better network weights to generate adversarial examples with better transferability,under fixed network architectures.Specifically,to reduce the model-specific features and obtain better output distributions,we construct a multi-teacher framework,where the knowledge is distilled from different teacher architectures into one student network.By considering that the gradient of input is usually utilized to generate adversarial examples,we impose constraints on the gradients between the student and teacher models,to further alleviate the output inconsistency problem and enhance the adversarial transferability.Extensive experiments demonstrate that our proposed work can significantly improve the adversarial transferability.展开更多
Magnetic Resonance Imaging(MRI)serves as a crucial diagnostic tool in medical practice,yet its lengthy acquisition times pose challenges for patient comfort and clinical efficiency.Current deep learning approaches for...Magnetic Resonance Imaging(MRI)serves as a crucial diagnostic tool in medical practice,yet its lengthy acquisition times pose challenges for patient comfort and clinical efficiency.Current deep learning approaches for accelerating MRI acquisition face difficulties in managing data variability caused by different scanner vendors or imaging protocols.This research investigates the use of transfer learning in variational deep learning models to enhance generalization capabilities.We collect 135 ACR phantom samples from 3.0T GE and SIEMENS MRI scanners,following standard ACR guidelines,to study vendor-specific generalization.Additionally,the fastMRI brain dataset,a recognized benchmark for MRI acceleration,is utilized to evaluate performance across diverse acquisition sequences.Through comprehensive testing,we identify vendor and sequence inconsistencies as key hurdles for accelerated MRI generalization.To overcome these challenges,we introduce a feature refinement-based transfer learning method,achieving significant gains over baseline models in both vendor and sequence generalization tasks.Moreover,we incorporate experience replay to mitigate catastrophic forgetting,resulting in notable performance stability.For vendor generalization,our approach reduces Peak Signal Noise-to-Ratio(PSNR)and Structural SIMilarity(SSIM)degradation by 25.55%and 9.5%,respectively.Similarly,for sequence transfer,forgetting is reduced by 3.5%(PSNR)and 2%(SSIM),establishing a robust framework with substantial improvements.展开更多
In this paper, aggregation question based on group decision making and a single decision making is studied. The theory of entropy is applied to the sets pair analysis. The system of relation entropy and the transferab...In this paper, aggregation question based on group decision making and a single decision making is studied. The theory of entropy is applied to the sets pair analysis. The system of relation entropy and the transferable entropy notion are put. The character is studied. An potential by the relation entropy and transferable entropy are defined. It is the consistency measure on the group between a single decision making. We gained a new aggregation effective definition on the group misjudge.展开更多
Cirrus clouds play a crucial role in the energy balance of the Earth-atmosphere system.We investigated the spatiotemporal variations of cirrus over the South China Sea(SCS)using satellite data(MOD08,MYD08,CALIPSO)and ...Cirrus clouds play a crucial role in the energy balance of the Earth-atmosphere system.We investigated the spatiotemporal variations of cirrus over the South China Sea(SCS)using satellite data(MOD08,MYD08,CALIPSO)and reanalysis data(MERRA-2)from March 2007 to February 2015(eight years).The horizontal distribution reveals lower cirrus fraction values in the northern SCS and higher values in the southern region,with minima observed in March and April and maxima sequentially occurring in August(northern SCS,NSCS),September(middle SCS,MSCS),and December(southern SCS,SSCS).Vertically,the cirrus fraction peaks in summer and reaches its lowest levels in spring.Opaque cirrus dominates during summer in the NSCS and MSCS,comprising 53.6%and 55.9%,respectively,while the SSCS exhibits a higher frequency of opaque cirrus relative to other cloud types.Subvisible cirrus clouds have the lowest frequency year-round,whereas thin cirrus is most prominent in winter in the NSCS(46.3%)and in spring in the MSCS(45.3%).A case study from September 2021 further explores the influence of ice crystal habits on brightness temperature(BT)over the SCS.Simulations utilizing five ice crystal shapes from the ARTS DDA(Atmospheric Radiative Transfer Simulator Discrete Dipole Approximation)database and the RTTOV 12.4 radiative transfer model reveal that the 8-column-aggregate shape best represents BT in the NSCS and SSCS,while the large-block-aggregate shape performs better in the SSCS.展开更多
Undesired ice accumulation on infrastructure and transportation systems leads to catastrophic events and significant economic losses.Although various anti-icing surfaces with photothermal effects can initially prevent...Undesired ice accumulation on infrastructure and transportation systems leads to catastrophic events and significant economic losses.Although various anti-icing surfaces with photothermal effects can initially prevent icing,any thawy droplets remaining on the horizontal surface can quickly re-freezing once the light diminishes.To address these challenges,we have developed a self-draining slippery surface(SDSS)that enables the thawy droplets to self-remove on the horizontal surface,thereby facilitating real-time anti-icing with the aid of sunlight(100 m W cm^(-2)).This is achieved by sandwiching a thin pyroelectric layer between slippery surface and photothermal film.Due to the synergy between the photothermal and pyroelectric layers,the SDSS not only maintains a high surface temperature of 19.8±2.2℃at the low temperature(-20.0±1.0℃),but also generates amount of charge through thermoelectric coupling.Thus,as cold droplets dropped on the SDSS,electrostatic force pushes the droplets off the charged surface because of the charge transfer mechanism.Even if the surface freezes overnight,the ice can melt and drain off the SDSS within 10 min of exposure to sunlight at-20.0±1.0℃,leaving a clean surface.This work provides a new perspective on the anti-icing system in the real-world environments.展开更多
Wide-temperature applications of sodium-ion batteries(SIBs)are severely limited by the sluggish ion insertion/diffusion kinetics of conversion-type anodes.Quantum-sized transition metal dichalcogenides possess unique ...Wide-temperature applications of sodium-ion batteries(SIBs)are severely limited by the sluggish ion insertion/diffusion kinetics of conversion-type anodes.Quantum-sized transition metal dichalcogenides possess unique advantages of charge delocalization and enrich uncoordinated electrons and short-range transfer kinetics,which are crucial to achieve rapid low-temperature charge transfer and high-temperature interface stability.Herein,a quantum-scale FeS_(2) loaded on three-dimensional Ti_(3)C_(2) MXene skeletons(FeS_(2) QD/MXene)fabricated as SIBs anode,demonstrating impressive performance under wide-temperature conditions(−35 to 65).The theoretical calculations combined with experimental characterization interprets that the unsaturated coordination edges of FeS_(2) QD can induce delocalized electronic regions,which reduces electrostatic potential and significantly facilitates efficient Na+diffusion across a broad temperature range.Moreover,the Ti_(3)C_(2) skeleton reinforces structural integrity via Fe-O-Ti bonding,while enabling excellent dispersion of FeS_(2) QD.As expected,FeS_(2) QD/MXene anode harvests capacities of 255.2 and 424.9 mAh g^(−1) at 0.1 A g^(−1) under−35 and 65,and the energy density of FeS_(2) QD/MXene//NVP full cell can reach to 162.4 Wh kg^(−1) at−35,highlighting its practical potential for wide-temperatures conditions.This work extends the uncoordinated regions induced by quantum-size effects for exceptional Na^(+)ion storage and diffusion performance at wide-temperatures environment.展开更多
Chemical exchange saturation transfer magnetic resonance imaging is an advanced imaging technique that enables the detection of compounds at low concentrations with high sensitivity and spatial resolution and has been...Chemical exchange saturation transfer magnetic resonance imaging is an advanced imaging technique that enables the detection of compounds at low concentrations with high sensitivity and spatial resolution and has been extensively studied for diagnosing malignancy and stroke.In recent years,the emerging exploration of chemical exchange saturation transfer magnetic resonance imaging for detecting pathological changes in neurodegenerative diseases has opened up new possibilities for early detection and repetitive scans without ionizing radiation.This review serves as an overview of chemical exchange saturation transfer magnetic resonance imaging with detailed information on contrast mechanisms and processing methods and summarizes recent developments in both clinical and preclinical studies of chemical exchange saturation transfer magnetic resonance imaging for Alzheimer’s disease,Parkinson’s disease,multiple sclerosis,and Huntington’s disease.A comprehensive literature search was conducted using databases such as PubMed and Google Scholar,focusing on peer-reviewed articles from the past 15 years relevant to clinical and preclinical applications.The findings suggest that chemical exchange saturation transfer magnetic resonance imaging has the potential to detect molecular changes and altered metabolism,which may aid in early diagnosis and assessment of the severity of neurodegenerative diseases.Although promising results have been observed in selected clinical and preclinical trials,further validations are needed to evaluate their clinical value.When combined with other imaging modalities and advanced analytical methods,chemical exchange saturation transfer magnetic resonance imaging shows potential as an in vivo biomarker,enhancing the understanding of neuropathological mechanisms in neurodegenerative diseases.展开更多
Free-form optoelectronic devices can provide hyper-connectivity over space and time.However,most conformable optoelectronic devices can only be fabricated on flat polymeric materials using low-temperature processes,li...Free-form optoelectronic devices can provide hyper-connectivity over space and time.However,most conformable optoelectronic devices can only be fabricated on flat polymeric materials using low-temperature processes,limiting their application and forms.This paper presents free-form optoelectronic devices that are not dependent on the shape or material.For medical applications,the transferable OLED(10μm)is formed in a sandwich structure with an ultra-thin transferable barrier(4.8μm).The results showed that the fabricated sandwich-structure transferable OLED(STOLED)exhibit the same high-efficiency performance on cylindricalshaped materials and on materials such as textile and paper.Because the neutral axis is freely adjustable using the sandwich structure,the textile-based OLED achieved both folding reliability and washing reliability,as well as a long operating life(>150 h).When keratinocytes were irradiated with red STOLED light,cell proliferation and cell migration increased by 26 and 32%,respectively.In the skin equivalent model,the epidermis thickness was increased by 39%;additionally,in organ culture,not only was the skin area increased by 14%,but also,reepithelialization was highly induced.Based on the results,the STOLED is expected to be applicable in various wearable and disposable photomedical devices.展开更多
The length and time scales of atomistic simulations are limited by the computational cost of the methods used to predict material properties.In recent years there has been great progress in the use of machine-learning...The length and time scales of atomistic simulations are limited by the computational cost of the methods used to predict material properties.In recent years there has been great progress in the use of machine-learning algorithms to develop fast and accurate interatomic potential models,but it remains a challenge to develop models that generalize well and are fast enough to be used at extreme time and length scales.To address this challenge,we have developed a machine-learning algorithm based on symbolic regression in the form of genetic programming that is capable of discovering accurate,computationally efficient many-body potential models.The key to our approach is to explore a hypothesis space of models based on fundamental physical principles and select models within this hypothesis space based on their accuracy,speed,and simplicity.The focus on simplicity reduces the risk of overfitting the training data and increases the chances of discovering a model that generalizes well.Our algorithm was validated by rediscovering an exact Lennard-Jones potential and a Sutton-Chen embedded-atom method potential from training data generated using these models.By using training data generated from density functional theory calculations,we found potential models for elemental copper that are simple,as fast as embedded-atom models,and capable of accurately predicting properties outside of their training set.Our approach requires relatively small sets of training data,making it possible to generate training data using highly accurate methods at a reasonable computational cost.We present our approach,the forms of the discovered models,and assessments of their transferability,accuracy and speed.展开更多
基金supported in part by the National Natural Science Foundation(62202118,U24A20241)in part by Major Scientific and Technological Special Project of Guizhou Province([2024]014,[2024]003)+1 种基金in part by Scientific and Technological Research Projects from Guizhou Education Department(Qian jiao ji[2023]003)in part by Guizhou Science and Technology Department Hundred Level Innovative Talents Project(GCC[2023]018).
文摘Deep neural networks remain susceptible to adversarial examples,where the goal of an adversarial attack is to introduce small perturbations to the original examples in order to confuse the model without being easily detected.Although many adversarial attack methods produce adversarial examples that have achieved great results in the whitebox setting,they exhibit low transferability in the black-box setting.In order to improve the transferability along the baseline of the gradient-based attack technique,we present a novel Stochastic Gradient Accumulation Momentum Iterative Attack(SAMI-FGSM)in this study.In particular,during each iteration,the gradient information is calculated using a normal sampling approach that randomly samples around the sample points,with the highest probability of capturing adversarial features.Meanwhile,the accumulated information of the sampled gradient from the previous iteration is further considered to modify the current updated gradient,and the original gradient attack direction is changed to ensure that the updated gradient direction is more stable.Comprehensive experiments conducted on the ImageNet dataset show that our method outperforms existing state-of-the-art gradient-based attack techniques,achieving an average improvement of 10.2%in transferability.
基金supported by the National Key Research and Development Program of China (2022YFB4602600)National Natural Science Foundation of China (Grant Nos. 52425508 & 52221001)the Hunan Provincial Natural Science Foundation of China (2025JJ60286)。
文摘Lithography is a Key enabling technique in modern micro/nano scale technology.Achieving the optimal trade-off between resolution,throughput,and cost remains a central focus in the ongoing development.However,current lithographic techniques such as direct-write,projection,and extreme ultraviolet lithography achieve higher resolution at the expense of increased complexity in optical systems or the use of shorter-wavelength light sources,thus raising the overall cost of production.Here,we present a cost-effective and wafer-level perfect conformal contact lithography at the diffraction limit.By leveraging a transferable photoresist,the technique ensures optimal contact between the mask and photoresist with zero-gap,facilitating the transfer of patterns at the diffraction limit while maintaining high fidelity and uniformity across large wafers.This technique applies to a wide range of complex surfaces,including non-conductive glass surfaces,flexible substrates,and curved surfaces.The proposed technique expands the potential of contact photolithography for novel device architectures and practic al manufacturing processes.
基金Project supported by the Southeast University Foundation for Excellent Young Scholars(No.4023001013)
文摘Whether the multi-biological toxicity from lead exposure could be transferred to progeny has not been clarified. In the present study, we explored the Caenorhabditis elegans to analyze the multiple toxicities from lead exposure and their possibly transferable properties. The lead exposure could cause series of severe multi-biological defects with a concentration-dependent manner by affecting the endpoints of life span, development, reproduction and locomotion behaviors in nematodes. Moreover, most of these toxicities could be transferred to progeny from lead exposed animals and some of the defects in progeny appeared even more severe than in their parents, such as the body sizes and mean life spans. We summarized the defects caused by lead exposure into three groups according to their transferable properties or rescue patterns. That is, the defects caused by lead exposure could be largely, or partially, or became even more severe in progeny animals. Therefore, our results suggest that lead exposure can cause severely multi-biological defects, and most of these multiple toxicities can be considered as transferable for exposed animals in C. elegans.
基金Project 49776300 supported by NSFC49925614 by the NSFC for Outstanding Young Scientists.
文摘Based on a new idea for research on cycling of marine biogenic elements, this study showed that only the leachable form phosphorus in natural grain sizes marine sediments constitutes the transferable phosphorous in the sediments. The transferable phosphorus content in the natural grain sizes surface sediments in the Huanghe River estuary adjacent waters ranges from 58.5-69.8 μg/g, accounting for only 9.1%-11.0% of the total phosphorus content, whereas the leachable form (“transferable") phosphorus content in the sediment after it was totally ground into powder was found to be 454.8-529.2 μg/g, accounting for 73.4%-89.1% of the total phosphorus. Analysis of the correlation between the biomass of benthos and the leachable form (“transferable") phosphorus showed that most of the leachable form (“transferable") phosphorus in the totally ground sediment did not participate in the marine biogeochemical cycling. Furthermore, a synchronous survey on benthos showed that the biomass of meio and macro benthos exhibited good positive correlation with the leachable form of phosphorus in the natural grain sizes sediment, but poorer correlation with the leachable form (“transferable") phosphorus in the totally ground sediment, indicating that transferable phosphorus in marine sediment is the leachable form of phosphorus in the natural grain sizes sediments, and is not the previously known leachable form (“transferable") phosphorus obtained from the totally ground sediment.
基金Funded by the Fundamental Research Funds for the Central Universities(No.DUT12ZD(G)01)the Opening Project of Key Laboratory of Inorganic Coating Materials,Chinese Academy of Sciences(No.KLICM-2012-01)
文摘ZnO thin films were deposited on graphite substrates by ultrasonic spray pyrolysis method with Zn（CH3COO）2·2H2O aqueous solution as precursor. The crystalline structure, morphology, and optical properties of the as-grown ZnO films were investigated systematically as a function of deposition temperature and growth time. Near-band edge ultraviolet （UV） emission was observed in room temperature photoluminescence spectra for the optimized samples, yet the usually observed defect related deep level emissions were nearly undetectable, indicating that high optical quality ZnO thin films could be achieved via this ultrasonic spray pyrolysis method. Considering the features of transferable and low thermal resistance of the graphite substrates, the achievement will be of special interest for the development of high-power semiconductor devices with sufficient vower durability.
基金the National Key R&D Program of China(2021ZD0112700,2018AAA0101400)the National Natural Science Foundation of China(62173251,61921004,U1713209)the Natural Science Foundation of Jiangsu Province of China(BK20202006)。
文摘In multi-agent reinforcement learning(MARL),the behaviors of each agent can influence the learning of others,and the agents have to search in an exponentially enlarged joint-action space.Hence,it is challenging for the multi-agent teams to explore in the environment.Agents may achieve suboptimal policies and fail to solve some complex tasks.To improve the exploring efficiency as well as the performance of MARL tasks,in this paper,we propose a new approach by transferring the knowledge across tasks.Differently from the traditional MARL algorithms,we first assume that the reward functions can be computed by linear combinations of a shared feature function and a set of taskspecific weights.Then,we define a set of basic MARL tasks in the source domain and pre-train them as the basic knowledge for further use.Finally,once the weights for target tasks are available,it will be easier to get a well-performed policy to explore in the target domain.Hence,the learning process of agents for target tasks is speeded up by taking full use of the basic knowledge that was learned previously.We evaluate the proposed algorithm on two challenging MARL tasks:cooperative boxpushing and non-monotonic predator-prey.The experiment results have demonstrated the improved performance compared with state-of-the-art MARL algorithms.
基金This work was supported by the National Natural Science Foundation of China(52275130)the National Key Research and Development Program of China(2018YFB1702400).
文摘Effective fault diagnosis of planetary gearboxes is critical for ensuring the safety and dependability of mechanical drive systems.Nevertheless,variable conditions and inadequate fault data bring huge challenges to its practical fault diagnosis.Taking this into account,this study presents a new intelligent fault diagnosis(IFD)approach for planetary gearbox using a transferable deep Q network(TDQN)that merges deep reinforcement learning(DRL)and transfer learning(TL).First,a DRL environment simulation is designed by a predefined classification Markov decision process.Then,leveraging varied-size convolutions and residual learning,a multiscale residual convolutional neural network agent for TDQN is created to automatically learn meaningful features directly from vibration signals while avoiding model degradation.Next,a large source dataset is obtained from complex conditions,and this agent learns an IFD policy via autonomous interaction with the data environment.Finally,a parameter-based TL strategy is adopted to retrain the model on target datasets with variable conditions and small training data,which is conducted by fine-tuning the model parameters gained from the source task to accomplish target tasks.The results show that this TDQN outperforms not only state-of-the-art methods in a source task with an accuracy of 98.53%but also in two target tasks with 99.63%and 98.37%,respectively.
基金supported by the National Natural Science Foundation of China(62333010,61673205).
文摘In the coal-to-ethylene glycol(CTEG)process,precisely estimating quality variables is crucial for process monitoring,optimization,and control.A significant challenge in this regard is relying on offline laboratory analysis to obtain these variables,which often incurs substantial monetary costs and significant time delays.The resulting few-shot learning scenarios present a hurdle to the efficient development of predictive models.To address this issue,our study introduces the transferable adversarial slow feature extraction network(TASF-Net),an innovative approach designed specifically for few-shot quality prediction in the CTEG process.TASF-Net uniquely integrates the slowness principle with a deep Bayesian framework,effectively capturing the nonlinear and inertial characteristics of the CTEG process.Additionally,the model employs a variable attention mechanism to identify quality-related input variables adaptively at each time step.A key strength of TASF-Net lies in its ability to navigate the complex measurement noise,outliers,and system interference typical in CTEG data.Adversarial learning strategy using a min-max game is adopted to improve its robustness and ability to model irregular industrial data accurately and significantly.Furthermore,an incremental refining transfer learning framework is designed to further improve few-shot prediction performance achieved by transferring knowledge from the pretrained model on the source domain to the target domain.The effectiveness and superiority of TASF-Net have been empirically validated using a real-world CTEG dataset.Compared with some state-of-the-art methods,TASF-Net demonstrates exceptional capability in addressing the intricate challenges for few-shot quality prediction in the CTEG process.
基金Supported by the Natural Science Foundation of Zhejiang Province, China (Y106802)Guangdong Provincial Science and Technology Planning Project of China (2007B010200035)
文摘In this paper, a new cash scheme is proposed for electronic payment system, in which the cash can be transferred several times. When this kind of cash is used, the fraud such as double spending can be found out but the bank and the trusted party needs not be involved online in each transaction. This cash system is anonymous in normal transactions. But if a fraud happens, the trusted party can withdraw the anonymity to find out the cheater. The new cash scheme is transferable, anonymous, off-line and efficient.
基金the National Natural Science Foundation of China under Grants U1836106 and 81961138010the Beijing Natural Science Foundation under Grants 19L2029 and M21032+2 种基金the Scientific and Technological Innovation Foundation of Foshan under Grants BK20BF010 and BK21BF001the Scientific and Technological Innovation Foundation of Shunde Graduate School,USTB,under Grant BK19BF006by the Fundamental Research Funds for the University of Science and Technology Beijing under Grant FRF-BD-19-012A.
文摘With the development of information technology,malware threats to the industrial system have become an emergent issue,since various industrial infrastructures have been deeply integrated into our modern works and lives.To identify and classify new malware variants,different types of deep learning models have been widely explored recently.Generally,sufficient data is usually required to achieve a well-trained deep learning classifier with satisfactory generalization ability.However,in current practical applications,an ample supply of data is absent in most specific industrial malware detection scenarios.Transfer learning as an effective approach can be used to alleviate the influence of the small sample size problem.In addition,it can also reuse the knowledge from pretrained models,which is beneficial to the real-time requirement in industrial malware detection.In this paper,we investigate the transferable features learned by a 1D-convolutional network and evaluate our proposed methods on 6 transfer learning tasks.The experiment results show that 1D-convolutional architecture is effective to learn transferable features for malware classification,and indicate that transferring the first 2 layers of our proposed 1D-convolutional network is the most efficient way to reuse the learned features.
基金supported in part by NSFC No.62202275 and Shandong-SF No.ZR2022QF012 projects.
文摘In recent years,deep learning(DL)models have achieved signifcant progress in many domains,such as autonomous driving,facial recognition,and speech recognition.However,the vulnerability of deep learning models to adversarial attacks has raised serious concerns in the community because of their insufcient robustness and generalization.Also,transferable attacks have become a prominent method for black-box attacks.In this work,we explore the potential factors that impact adversarial examples(AEs)transferability in DL-based speech recognition.We also discuss the vulnerability of diferent DL systems and the irregular nature of decision boundaries.Our results show a remarkable diference in the transferability of AEs between speech and images,with the data relevance being low in images but opposite in speech recognition.Motivated by dropout-based ensemble approaches,we propose random gradient ensembles and dynamic gradient-weighted ensembles,and we evaluate the impact of ensembles on the transferability of AEs.The results show that the AEs created by both approaches are valid for transfer to the black box API.
基金supported in part by the National Natural Science Foundation of China(Grant Nos.62272020 and U20B2069)in part by the State Key Laboratory of Complex&Critical Software Environment(SKLSDE2023ZX-16)in part by the Fundamental Research Funds for Central Universities.
文摘This paper focuses on an important type of black-box attacks,i.e.,transfer-based adversarial attacks,where the adversary generates adversarial examples using a substitute(source)model and utilizes them to attack an unseen target model,without knowing its information.Existing methods tend to give unsatisfactory adversarial transferability when the source and target models are from different types of DNN architectures(e.g.,ResNet-18 and Swin Transformer).In this paper,we observe that the above phenomenon is induced by the output inconsistency problem.To alleviate this problem while effectively utilizing the existing DNN models,we propose a common knowledge learning(CKL)framework to learn better network weights to generate adversarial examples with better transferability,under fixed network architectures.Specifically,to reduce the model-specific features and obtain better output distributions,we construct a multi-teacher framework,where the knowledge is distilled from different teacher architectures into one student network.By considering that the gradient of input is usually utilized to generate adversarial examples,we impose constraints on the gradients between the student and teacher models,to further alleviate the output inconsistency problem and enhance the adversarial transferability.Extensive experiments demonstrate that our proposed work can significantly improve the adversarial transferability.
基金supported by the following grants:the National Institutes of Health(Nos.RF1AG073424 and P30AG072980)the Arizona Department of Health Services(No.CTR057001).
文摘Magnetic Resonance Imaging(MRI)serves as a crucial diagnostic tool in medical practice,yet its lengthy acquisition times pose challenges for patient comfort and clinical efficiency.Current deep learning approaches for accelerating MRI acquisition face difficulties in managing data variability caused by different scanner vendors or imaging protocols.This research investigates the use of transfer learning in variational deep learning models to enhance generalization capabilities.We collect 135 ACR phantom samples from 3.0T GE and SIEMENS MRI scanners,following standard ACR guidelines,to study vendor-specific generalization.Additionally,the fastMRI brain dataset,a recognized benchmark for MRI acceleration,is utilized to evaluate performance across diverse acquisition sequences.Through comprehensive testing,we identify vendor and sequence inconsistencies as key hurdles for accelerated MRI generalization.To overcome these challenges,we introduce a feature refinement-based transfer learning method,achieving significant gains over baseline models in both vendor and sequence generalization tasks.Moreover,we incorporate experience replay to mitigate catastrophic forgetting,resulting in notable performance stability.For vendor generalization,our approach reduces Peak Signal Noise-to-Ratio(PSNR)and Structural SIMilarity(SSIM)degradation by 25.55%and 9.5%,respectively.Similarly,for sequence transfer,forgetting is reduced by 3.5%(PSNR)and 2%(SSIM),establishing a robust framework with substantial improvements.
基金This projectis supported by National Natural Science Foundation of China( No.79930 90 0 )
文摘In this paper, aggregation question based on group decision making and a single decision making is studied. The theory of entropy is applied to the sets pair analysis. The system of relation entropy and the transferable entropy notion are put. The character is studied. An potential by the relation entropy and transferable entropy are defined. It is the consistency measure on the group between a single decision making. We gained a new aggregation effective definition on the group misjudge.
基金supported by the National Natural Science Foundation of China(Grant Nos.42027804,41775026,and 41075012)。
文摘Cirrus clouds play a crucial role in the energy balance of the Earth-atmosphere system.We investigated the spatiotemporal variations of cirrus over the South China Sea(SCS)using satellite data(MOD08,MYD08,CALIPSO)and reanalysis data(MERRA-2)from March 2007 to February 2015(eight years).The horizontal distribution reveals lower cirrus fraction values in the northern SCS and higher values in the southern region,with minima observed in March and April and maxima sequentially occurring in August(northern SCS,NSCS),September(middle SCS,MSCS),and December(southern SCS,SSCS).Vertically,the cirrus fraction peaks in summer and reaches its lowest levels in spring.Opaque cirrus dominates during summer in the NSCS and MSCS,comprising 53.6%and 55.9%,respectively,while the SSCS exhibits a higher frequency of opaque cirrus relative to other cloud types.Subvisible cirrus clouds have the lowest frequency year-round,whereas thin cirrus is most prominent in winter in the NSCS(46.3%)and in spring in the MSCS(45.3%).A case study from September 2021 further explores the influence of ice crystal habits on brightness temperature(BT)over the SCS.Simulations utilizing five ice crystal shapes from the ARTS DDA(Atmospheric Radiative Transfer Simulator Discrete Dipole Approximation)database and the RTTOV 12.4 radiative transfer model reveal that the 8-column-aggregate shape best represents BT in the NSCS and SSCS,while the large-block-aggregate shape performs better in the SSCS.
基金supported by the National Natural Science Foundation of China(52273101,51922018,and 21875011)the Fundamental Research Funds for the Central Universities(KG21015201 and KG21020801)China Postdoctoral Science Foundation(2025M77422)。
文摘Undesired ice accumulation on infrastructure and transportation systems leads to catastrophic events and significant economic losses.Although various anti-icing surfaces with photothermal effects can initially prevent icing,any thawy droplets remaining on the horizontal surface can quickly re-freezing once the light diminishes.To address these challenges,we have developed a self-draining slippery surface(SDSS)that enables the thawy droplets to self-remove on the horizontal surface,thereby facilitating real-time anti-icing with the aid of sunlight(100 m W cm^(-2)).This is achieved by sandwiching a thin pyroelectric layer between slippery surface and photothermal film.Due to the synergy between the photothermal and pyroelectric layers,the SDSS not only maintains a high surface temperature of 19.8±2.2℃at the low temperature(-20.0±1.0℃),but also generates amount of charge through thermoelectric coupling.Thus,as cold droplets dropped on the SDSS,electrostatic force pushes the droplets off the charged surface because of the charge transfer mechanism.Even if the surface freezes overnight,the ice can melt and drain off the SDSS within 10 min of exposure to sunlight at-20.0±1.0℃,leaving a clean surface.This work provides a new perspective on the anti-icing system in the real-world environments.
基金supported by the National Nature Science Foundation of China(Nos.52202335 and 52171227)Natural Science Foundation of Jiangsu Province(No.BK20221137)National Key R&D Program of China(2024YFE0108500).
文摘Wide-temperature applications of sodium-ion batteries(SIBs)are severely limited by the sluggish ion insertion/diffusion kinetics of conversion-type anodes.Quantum-sized transition metal dichalcogenides possess unique advantages of charge delocalization and enrich uncoordinated electrons and short-range transfer kinetics,which are crucial to achieve rapid low-temperature charge transfer and high-temperature interface stability.Herein,a quantum-scale FeS_(2) loaded on three-dimensional Ti_(3)C_(2) MXene skeletons(FeS_(2) QD/MXene)fabricated as SIBs anode,demonstrating impressive performance under wide-temperature conditions(−35 to 65).The theoretical calculations combined with experimental characterization interprets that the unsaturated coordination edges of FeS_(2) QD can induce delocalized electronic regions,which reduces electrostatic potential and significantly facilitates efficient Na+diffusion across a broad temperature range.Moreover,the Ti_(3)C_(2) skeleton reinforces structural integrity via Fe-O-Ti bonding,while enabling excellent dispersion of FeS_(2) QD.As expected,FeS_(2) QD/MXene anode harvests capacities of 255.2 and 424.9 mAh g^(−1) at 0.1 A g^(−1) under−35 and 65,and the energy density of FeS_(2) QD/MXene//NVP full cell can reach to 162.4 Wh kg^(−1) at−35,highlighting its practical potential for wide-temperatures conditions.This work extends the uncoordinated regions induced by quantum-size effects for exceptional Na^(+)ion storage and diffusion performance at wide-temperatures environment.
基金supported by The University of Hong Kong,China(109000487,109001694,204610401,and 204610519)National Natural Science Foundation of China(82402225)(to JH).
文摘Chemical exchange saturation transfer magnetic resonance imaging is an advanced imaging technique that enables the detection of compounds at low concentrations with high sensitivity and spatial resolution and has been extensively studied for diagnosing malignancy and stroke.In recent years,the emerging exploration of chemical exchange saturation transfer magnetic resonance imaging for detecting pathological changes in neurodegenerative diseases has opened up new possibilities for early detection and repetitive scans without ionizing radiation.This review serves as an overview of chemical exchange saturation transfer magnetic resonance imaging with detailed information on contrast mechanisms and processing methods and summarizes recent developments in both clinical and preclinical studies of chemical exchange saturation transfer magnetic resonance imaging for Alzheimer’s disease,Parkinson’s disease,multiple sclerosis,and Huntington’s disease.A comprehensive literature search was conducted using databases such as PubMed and Google Scholar,focusing on peer-reviewed articles from the past 15 years relevant to clinical and preclinical applications.The findings suggest that chemical exchange saturation transfer magnetic resonance imaging has the potential to detect molecular changes and altered metabolism,which may aid in early diagnosis and assessment of the severity of neurodegenerative diseases.Although promising results have been observed in selected clinical and preclinical trials,further validations are needed to evaluate their clinical value.When combined with other imaging modalities and advanced analytical methods,chemical exchange saturation transfer magnetic resonance imaging shows potential as an in vivo biomarker,enhancing the understanding of neuropathological mechanisms in neurodegenerative diseases.
基金supported by the Engineering Research Center of Excellence(ERC)Programsupported by the National Research Foundation(NRF)of the Korean Ministry of ScienceICT&Future Planning(MSIP)(Grant No.NRF-2017R1A5A1014708).
文摘Free-form optoelectronic devices can provide hyper-connectivity over space and time.However,most conformable optoelectronic devices can only be fabricated on flat polymeric materials using low-temperature processes,limiting their application and forms.This paper presents free-form optoelectronic devices that are not dependent on the shape or material.For medical applications,the transferable OLED(10μm)is formed in a sandwich structure with an ultra-thin transferable barrier(4.8μm).The results showed that the fabricated sandwich-structure transferable OLED(STOLED)exhibit the same high-efficiency performance on cylindricalshaped materials and on materials such as textile and paper.Because the neutral axis is freely adjustable using the sandwich structure,the textile-based OLED achieved both folding reliability and washing reliability,as well as a long operating life(>150 h).When keratinocytes were irradiated with red STOLED light,cell proliferation and cell migration increased by 26 and 32%,respectively.In the skin equivalent model,the epidermis thickness was increased by 39%;additionally,in organ culture,not only was the skin area increased by 14%,but also,reepithelialization was highly induced.Based on the results,the STOLED is expected to be applicable in various wearable and disposable photomedical devices.
基金We acknowledge financial support from the Office of Naval Research,grant number N000141512665.
文摘The length and time scales of atomistic simulations are limited by the computational cost of the methods used to predict material properties.In recent years there has been great progress in the use of machine-learning algorithms to develop fast and accurate interatomic potential models,but it remains a challenge to develop models that generalize well and are fast enough to be used at extreme time and length scales.To address this challenge,we have developed a machine-learning algorithm based on symbolic regression in the form of genetic programming that is capable of discovering accurate,computationally efficient many-body potential models.The key to our approach is to explore a hypothesis space of models based on fundamental physical principles and select models within this hypothesis space based on their accuracy,speed,and simplicity.The focus on simplicity reduces the risk of overfitting the training data and increases the chances of discovering a model that generalizes well.Our algorithm was validated by rediscovering an exact Lennard-Jones potential and a Sutton-Chen embedded-atom method potential from training data generated using these models.By using training data generated from density functional theory calculations,we found potential models for elemental copper that are simple,as fast as embedded-atom models,and capable of accurately predicting properties outside of their training set.Our approach requires relatively small sets of training data,making it possible to generate training data using highly accurate methods at a reasonable computational cost.We present our approach,the forms of the discovered models,and assessments of their transferability,accuracy and speed.
