Generative image steganography is a technique that directly generates stego images from secret infor-mation.Unlike traditional methods,it theoretically resists steganalysis because there is no cover image.Currently,th...Generative image steganography is a technique that directly generates stego images from secret infor-mation.Unlike traditional methods,it theoretically resists steganalysis because there is no cover image.Currently,the existing generative image steganography methods generally have good steganography performance,but there is still potential room for enhancing both the quality of stego images and the accuracy of secret information extraction.Therefore,this paper proposes a generative image steganography algorithm based on attribute feature transformation and invertible mapping rule.Firstly,the reference image is disentangled by a content and an attribute encoder to obtain content features and attribute features,respectively.Then,a mean mapping rule is introduced to map the binary secret information into a noise vector,conforming to the distribution of attribute features.This noise vector is input into the generator to produce the attribute transformed stego image with the content feature of the reference image.Additionally,we design an adversarial loss,a reconstruction loss,and an image diversity loss to train the proposed model.Experimental results demonstrate that the stego images generated by the proposed method are of high quality,with an average extraction accuracy of 99.4%for the hidden information.Furthermore,since the stego image has a uniform distribution similar to the attribute-transformed image without secret information,it effectively resists both subjective and objective steganalysis.展开更多
Deep neural networks(DNNs)and generative AI(GenAI)are increasingly vulnerable to backdoor attacks,where adversaries embed triggers into inputs to cause models to misclassify or misinterpret target labels.Beyond tradit...Deep neural networks(DNNs)and generative AI(GenAI)are increasingly vulnerable to backdoor attacks,where adversaries embed triggers into inputs to cause models to misclassify or misinterpret target labels.Beyond traditional single-trigger scenarios,attackers may inject multiple triggers across various object classes,forming unseen backdoor-object configurations that evade standard detection pipelines.In this paper,we introduce DBOM(Disentangled Backdoor-Object Modeling),a proactive framework that leverages structured disentanglement to identify and neutralize both seen and unseen backdoor threats at the dataset level.Specifically,DBOM factorizes input image representations by modeling triggers and objects as independent primitives in the embedding space through the use of Vision-Language Models(VLMs).By leveraging the frozen,pre-trained encoders of VLMs,our approach decomposes the latent representations into distinct components through a learnable visual prompt repository and prompt prefix tuning,ensuring that the relationships between triggers and objects are explicitly captured.To separate trigger and object representations in the visual prompt repository,we introduce the trigger–object separation and diversity losses that aids in disentangling trigger and object visual features.Next,by aligning image features with feature decomposition and fusion,as well as learned contextual prompt tokens in a shared multimodal space,DBOM enables zero-shot generalization to novel trigger-object pairings that were unseen during training,thereby offering deeper insights into adversarial attack patterns.Experimental results on CIFAR-10 and GTSRB demonstrate that DBOM robustly detects poisoned images prior to downstream training,significantly enhancing the security of DNN training pipelines.展开更多
Learning disentangled representation of data is a key problem in deep learning.Specifically,disentangling 2D facial landmarks into different factors(e.g.,identity and expression)is widely used in the applications of f...Learning disentangled representation of data is a key problem in deep learning.Specifically,disentangling 2D facial landmarks into different factors(e.g.,identity and expression)is widely used in the applications of face reconstruction,face reenactment and talking head et al..However,due to the sparsity of landmarks and the lack of accurate labels for the factors,it is hard to learn the disentangled representation of landmarks.To address these problem,we propose a simple and effective model named FLD-VAE to disentangle arbitrary facial landmarks into identity and expression latent representations,which is based on a Variational Autoencoder framework.Besides,we propose three invariant loss functions in both latent and data levels to constrain the invariance of representations during training stage.Moreover,we implement an identity preservation loss to further enhance the representation ability of identity factor.To the best of our knowledge,this is the first work to end-to-end disentangle identity and expression factors simultaneously from one single facial landmark.展开更多
1 Introduction Molecular Property Prediction aims to identify molecules sharing similar efficacious properties[1],which is a foundational task in drug discovery,materials science and bioinformatics.Graph neural networ...1 Introduction Molecular Property Prediction aims to identify molecules sharing similar efficacious properties[1],which is a foundational task in drug discovery,materials science and bioinformatics.Graph neural networks(GNNs)have shown significant success in this field.However,GNN-based methods often face label scarcity,limiting their performance in predicting molecular properties.Besides,GNNs trained on specific datasets frequently struggle with generalization due to domain shift[2].展开更多
Currently,most existing inductive relation prediction approaches are based on subgraph structures,with subgraph features extracted using graph neural networks to predict relations.However,subgraphs may contain disconn...Currently,most existing inductive relation prediction approaches are based on subgraph structures,with subgraph features extracted using graph neural networks to predict relations.However,subgraphs may contain disconnected regions,which usually represent different semantic ranges.Because not all semantic information about the regions is helpful in relation prediction,we propose a relation prediction model based on a disentangled subgraph structure and implement a feature updating approach based on relevant semantic aggregation.To indirectly achieve the disentangled subgraph structure from a semantic perspective,the mapping of entity features into different semantic spaces and the aggregation of related semantics on each semantic space are updated.The disentangled model can focus on features having higher semantic relevance in the prediction,thus addressing a problem with existing approaches,which ignore the semantic differences in different subgraph structures.Furthermore,using a gated recurrent neural network,this model enhances the features of entities by sorting them by distance and extracting the path information in the subgraphs.Experimentally,it is shown that when there are numerous disconnected regions in the subgraph,our model outperforms existing mainstream models in terms of both Area Under the Curve-Precision-Recall(AUC-PR)and Hits@10.Experiments prove that semantic differences in the knowledge graph can be effectively distinguished and verify the effectiveness of this method.展开更多
In recent years,the concept of digital human has attracted widespread attention from all walks of life,and the modelling of high-fidelity human bodies,heads,and hands has been intensively studied.This paper focuses on...In recent years,the concept of digital human has attracted widespread attention from all walks of life,and the modelling of high-fidelity human bodies,heads,and hands has been intensively studied.This paper focuses on head modelling and proposes a generic head parametric model based on neural radiance fields.Specifically,we first use face recognition networks and 3D facial expression database FaceWarehouse to parameterize identity and expression semantics,respectively,and use both as conditional inputs to build a neural radiance field for the human head,thereby improving the head model’s representation ability while ensuring editing capabilities for the identity and expression of the rendered results;then,through a combination of volume rendering and neural rendering,the 3D representation of the head is rapidly rendered into the 2D plane,producing a high-fidelity image of the human head.Thanks to the well-designed loss functions and good implicit representation of the neural radiance field,our model can not only edit the identity and expression independently,but also freely modify the virtual camera position of the rendering results.It has excellent multi-view consistency,and has many applications in novel view synthesis,pose driving and more.展开更多
Deep neural networks are known to be vulnerable to adversarial attacks.Unfortunately,the underlying mechanisms remain insufficiently understood,leading to empirical defenses that often fail against new attacks.In this...Deep neural networks are known to be vulnerable to adversarial attacks.Unfortunately,the underlying mechanisms remain insufficiently understood,leading to empirical defenses that often fail against new attacks.In this paper,we explain adversarial attacks from the perspective of robust features,and propose a novel Generative Adversarial Network(GAN)-based Robust Feature Disentanglement framework(GRFD)for adversarial defense.The core of GRFD is an adversarial disentanglement structure comprising a generator and a discriminator.For the generator,we introduce a novel Latent Variable Constrained Variational Auto-Encoder(LVCVAE),which enhances the typical beta-VAE with a constrained rectification module to enforce explicit clustering of latent variables.To supervise the disentanglement of robust features,we design a Robust Supervisory Model(RSM)as the discriminator,sharing architectural alignment with the target model.The key innovation of RSM is our proposed Feature Robustness Metric(FRM),which serves as part of the training loss and synthesizes the classification ability of features as well as their resistance to perturbations.Extensive experiments on three benchmark datasets demonstrate the superiority of GRFD:it achieves 93.69%adversarial accuracy on MNIST,77.21%on CIFAR10,and 58.91%on CIFAR100 with minimal degradation in clean accuracy.Codes are available at:(accessed on 23 July 2025).展开更多
Background Numerous studies have consistently demonstrated that a considerable proportion of patients with major depressive disorder (MDD) frequently exhibit pronounced dyslipidaemia. However, the causal dynamics betw...Background Numerous studies have consistently demonstrated that a considerable proportion of patients with major depressive disorder (MDD) frequently exhibit pronounced dyslipidaemia. However, the causal dynamics between MDD and dyslipidaemia remain elusive.Aims To comprehensively disentangle the genetic causality between MDD and various phenotypes of blood lipids, thereby facilitating the advancement of management strategies for these conditions.Methods We conducted a two-sample univariable Mendelian randomisation (MR) analysis using different models, including the inverse variance weighted (IVW) method and causal analysis using the summary effect (CAUSE) estimates, as well as a multivariable MR analysis. This analysis used summary statistics from genome-wide association studies (GWAS) of MDD and five lipid traits: low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, non-high-density lipoprotein cholesterol, total cholesterol and triglycerides (TG), encompassing 5 237 893 individuals of European and East Asian ancestries. For MDD, a total of 598 701 individuals were included, with 500 199 individuals of European ancestry (Ncase=170 756, Ncontrol=329 443) and 98 502 of East Asian ancestry (Ncase=12 588, Ncontrol=85 914). Lipid data were collected from 4 639 192 individuals through the Global Lipids Genetics Consortium (European, N=4 096 085;East Asian, N=543 107). Next, we used the two-step MR to explore the mediating factors between MDD and TG, and the risk factors affecting TG through MDD. Finally, we conducted a GWAS meta-analysis and enrichment analysis.Results In univariable MR, we observed a negative causal effect of low-density lipoprotein on MDD in both European populations (IVW: odds ratio (OR): 0.972, 95% confidence interval (CI) 0.947 to 0.998, p=0.037) and East Asian populations (IVW: OR: 0.928, 95% CI 0.864 to 0.997, p=0.042). Additionally, we identified a bidirectional causal relationship between TG and MDD, with TG having a causal effect on MDD (IVW: OR: 1.052, 95% CI 1.020 to 1.085, p=0.001) and MDD having a causal effect on TG (IVW: OR: 1.075, 95% CI 1.047 to 1.104, p<0.001). Multivariable MR analysis further supported the role of TG in MDD (OR: 1.205, 95% CI 1.034 to 1.405, p=0.017). CAUSE estimates indicated that the causal model of MDD on TG provided a better fit than the sharing model (p=0.003), while the association of TG on MDD was more likely due to horizontal correlated pleiotropy than causality. Mediation analyses revealed that waist-hip ratio (WHR) mediated 69% of the total causal effect of MDD on TG, while other identified risk factors exhibited lower mediating proportions either mediated through MDD (≤17%) or originating from MDD (≤29%). The GWAS meta-analysis highlighted potential pathways related to lipid processes and nucleosome assembling, with significant cell types identified in brain regions and liver tissues.Conclusions The findings indicate that genetic proxies of MDD are associated with elevated levels of TG, with WHR serving as a clinical indicator of the association. This suggests that interventions targeting WHR may be effective in reducing TG levels in patients with MDD.展开更多
representation that can identify and isolate different potential variables hidden in the highdimensional observations.Disentangled representation learning can capture information about a single change factor and contr...representation that can identify and isolate different potential variables hidden in the highdimensional observations.Disentangled representation learning can capture information about a single change factor and control it by the corresponding potential subspace,providing a robust representation for complex changes in the data.In this paper,we first introduce and analyze the current status of research on disentangled representation and its causal mechanisms and summarize three crucial properties of disentangled representation.Then,disentangled representation learning algorithms are classified into four categories and outlined in terms of both mathematical description and applicability.Subsequently,the loss functions and objective evaluation metrics commonly used in existing work on disentangled representation are classified.Finally,the paper summarizes representative applications of disentangled representation learning in the field of remote sensing and discusses its future development.展开更多
Using molecular dynamics(MD)simulations,this study explores the fluid properties of three polymer melts with the same number of entanglements,Z,achieved by adjusting the entanglement length Ne,while investigating the ...Using molecular dynamics(MD)simulations,this study explores the fluid properties of three polymer melts with the same number of entanglements,Z,achieved by adjusting the entanglement length Ne,while investigating the evolution of polymer melt conformation and entanglement under high-rate elongational flow.The identification of a master curve indicates consistent normalized linear viscoelastic behavior.Surprising findings regarding the steady-state viscosity at various elongational rates(Wi_(R)>4.7)for polymer melts with the same Z have been uncovered,challenging existing tube models.Nevertheless,the study demonstrates the potential for normalizing the steady-state elongational viscosity at high rates(Wi_(R)>4.7)by scaling with the square of the chain contour length.Additionally,the observed independence of viscosity on the elongational rate at high rates suggests that higher rates lead to a more significant alignment of polymer chains,a decrease in entanglement,and a stretching in contour length of polymer chains.Molecular-level tracking of tagged chains further supports the assumption of no entanglement under rapid elongation,emphasizing the need for further research on disentanglement in polymer melts subjected to high-rate elongational flow.These results carry significant implications for understanding and predicting the behavior of polymer melts under high-rate elongational flow conditions.展开更多
Tailoring the electronic metal-support interaction(EMSI)has attracted considerable interests as one of the most efficient approaches to improve both the activity and stability of metal catalysts in heterogeneous catal...Tailoring the electronic metal-support interaction(EMSI)has attracted considerable interests as one of the most efficient approaches to improve both the activity and stability of metal catalysts in heterogeneous catalysis.In this viewpoint,we illustrate the methodology and relevant fundamentals on the disentanglement,characterization,and interpretation of EMSI.Under the choice of monometallic catalyst over inert support,a combination of optimal experiment design,multimodal techniques,in situ characterization,with a comprehensive understanding of the underlying measurement protocols is highly desirable for a reliable determination of EMSI.Accordingly,not only the d-band filling but also d-band energy within the EMSI should be taken into consideration for providing general principles to guide the electron-promoting catalytic reaction.展开更多
By taking into account spatial degrees of freedom of atoms, we study the internal-state disentanglement dynamics of two atoms interacting with a vacuum multi-mode noise field. We show that the complete internal-state ...By taking into account spatial degrees of freedom of atoms, we study the internal-state disentanglement dynamics of two atoms interacting with a vacuum multi-mode noise field. We show that the complete internal-state disentanglement of the two atoms, caused due to the atomic spontaneous emission can be achieved in a finite time.展开更多
The bilinear generating function for products of two Laguerre 2D polynomials with different arguments is calculated. It corresponds to the formula of Mehler for the generating function of products of two Hermite polyn...The bilinear generating function for products of two Laguerre 2D polynomials with different arguments is calculated. It corresponds to the formula of Mehler for the generating function of products of two Hermite polynomials. Furthermore, the generating function for mixed products of Laguerre 2D and Hermite 2D polynomials and for products of two Hermite 2D polynomials is calculated. A set of infinite sums over products of two Laguerre 2D polynomials as intermediate step to the generating function for products of Laguerre 2D polynomials is evaluated but these sums possess also proper importance for calculations with Laguerre polynomials. With the technique of operator disentanglement some operator identities are derived in an appendix. They allow calculating convolutions of Gaussian functions combined with polynomials in one- and two-dimensional case and are applied to evaluate the discussed generating functions.展开更多
We explore the theoretical possibility of extending the usual squeezed state to those produced by nonlinear singlemode squeezing operators. We derive the wave functions of exp[-（ig/2）（（1-X2）1/2P ＋ P（1-X2）1/2）...We explore the theoretical possibility of extending the usual squeezed state to those produced by nonlinear singlemode squeezing operators. We derive the wave functions of exp[-（ig/2）（（1-X2）1/2P ＋ P（1-X2）1/2）]｜0 in the coordinate representation. A new operator＇s disentangling formula is derived as a by-product.展开更多
We recommend a new convenient method for disentangling some exponential operators and derive a set of new operator identities. Especially, we derive the normal odering form of exp [fa^＋a ＋ ga^2＋ ＋ ka^2] without ap...We recommend a new convenient method for disentangling some exponential operators and derive a set of new operator identities. Especially, we derive the normal odering form of exp [fa^＋a ＋ ga^2＋ ＋ ka^2] without appealing to Lie algebra method. Application of these formulas in solving some dynamic Hamiltonian is presented.展开更多
The disentanglement evolution of bipartite spin-1/2 system coupled to a common surrounding XY chain in transverse fields at nonzero temperature is studied in this letter. The dynamical process of the entanglement is n...The disentanglement evolution of bipartite spin-1/2 system coupled to a common surrounding XY chain in transverse fields at nonzero temperature is studied in this letter. The dynamical process of the entanglement is numerically and analytically investigated. We find that thermal effects can enhance disentanglement if the entangled initial state of the central spins does not in the decoherence free space. The critical phenomenon of quantum phase transitions reflected in the disentanglement can be washed out by the thermal effect eventually.展开更多
In the preceding paper [Commun. Theor. Phys. 51 （2009） 321] we have recommended a convenient method for disentangling exponential operators in the form of exp{B ＋ C}, trying to find an operator A that satisfies [A,...In the preceding paper [Commun. Theor. Phys. 51 （2009） 321] we have recommended a convenient method for disentangling exponential operators in the form of exp{B ＋ C}, trying to find an operator A that satisfies [A, B] = C, and [A, [A, B]] = 0, then from the Baker-Hausdorff formula we have exp{B ＋C} ： exp（B ＋ [A, B]} = e^A e^B e^-A. After arranging e^Ae^B = e^B e^A e^W, the disentangling exp{B ＋ C} = e^B e^W is obtained. In this work we use this method to two-mode case, especially, derive the normal ordering form of exp[h（a^＋a ＋ b^＋b） ＋ ga^＋b^＋ ＋ kab] without appealing to Lie algebra method.展开更多
We investigate the entanglement evolution of two qubits that are initially in Werner state under the classical phase noise. We discuss the influence of mixture degree on disentanglement. It is showed that the more mix...We investigate the entanglement evolution of two qubits that are initially in Werner state under the classical phase noise. We discuss the influence of mixture degree on disentanglement. It is showed that the more mixed the state, the shorter is the time of disentanglement.展开更多
In this article,two new algorithms are presented that convert a given data tensor train into either a Tucker decomposition with orthogonal matrix factors or a multi-scale entanglement renormalization ansatz(MERA).The ...In this article,two new algorithms are presented that convert a given data tensor train into either a Tucker decomposition with orthogonal matrix factors or a multi-scale entanglement renormalization ansatz(MERA).The Tucker core tensor is never explicitly computed but stored as a tensor train instead,resulting in both computationally and storage efficient algorithms.Both the multilinear Tucker-ranks as well as the MERA-ranks are automatically determined by the algorithm for a given upper bound on the relative approximation error.In addition,an iterative algorithm with low computational complexity based on solving an orthogonal Procrustes problem is proposed for the first time to retrieve optimal rank-lowering disentangler tensors,which are a crucial component in the construction of a low-rank MERA.Numerical experiments demonstrate the effectiveness of the proposed algorithms together with the potential storage benefit of a low-rank MERA over a tensor train.展开更多
A straightforward simple proof is given that dark energy is the natural conse-quence of a quantum disentanglement physical process. Thus while the ordinary energy density of the cosmos is equal to half that of Hardy’...A straightforward simple proof is given that dark energy is the natural conse-quence of a quantum disentanglement physical process. Thus while the ordinary energy density of the cosmos is equal to half that of Hardy’s quantum probability of Entanglement i.e. where , the density of cosmic dark energy is consequently one minus divided by two i.e. . This result is in full agreement with all the numerous previous theoretical predictions as well as being in remarkable agreement with the overwhelming majority of cosmic accurate measurements and observations.展开更多
基金supported in part by the National Natural Science Foundation of China(Nos.62202234,62401270)the China Postdoctoral Science Foundation(No.2023M741778)the Natural Science Foundation of Jiangsu Province(Nos.BK20240706,BK20240694).
文摘Generative image steganography is a technique that directly generates stego images from secret infor-mation.Unlike traditional methods,it theoretically resists steganalysis because there is no cover image.Currently,the existing generative image steganography methods generally have good steganography performance,but there is still potential room for enhancing both the quality of stego images and the accuracy of secret information extraction.Therefore,this paper proposes a generative image steganography algorithm based on attribute feature transformation and invertible mapping rule.Firstly,the reference image is disentangled by a content and an attribute encoder to obtain content features and attribute features,respectively.Then,a mean mapping rule is introduced to map the binary secret information into a noise vector,conforming to the distribution of attribute features.This noise vector is input into the generator to produce the attribute transformed stego image with the content feature of the reference image.Additionally,we design an adversarial loss,a reconstruction loss,and an image diversity loss to train the proposed model.Experimental results demonstrate that the stego images generated by the proposed method are of high quality,with an average extraction accuracy of 99.4%for the hidden information.Furthermore,since the stego image has a uniform distribution similar to the attribute-transformed image without secret information,it effectively resists both subjective and objective steganalysis.
基金supported by the UWF Argo Cyber Emerging Scholars(ACES)program funded by the National Science Foundation(NSF)CyberCorps^(®) Scholarship for Service(SFS)award under grant number 1946442.
文摘Deep neural networks(DNNs)and generative AI(GenAI)are increasingly vulnerable to backdoor attacks,where adversaries embed triggers into inputs to cause models to misclassify or misinterpret target labels.Beyond traditional single-trigger scenarios,attackers may inject multiple triggers across various object classes,forming unseen backdoor-object configurations that evade standard detection pipelines.In this paper,we introduce DBOM(Disentangled Backdoor-Object Modeling),a proactive framework that leverages structured disentanglement to identify and neutralize both seen and unseen backdoor threats at the dataset level.Specifically,DBOM factorizes input image representations by modeling triggers and objects as independent primitives in the embedding space through the use of Vision-Language Models(VLMs).By leveraging the frozen,pre-trained encoders of VLMs,our approach decomposes the latent representations into distinct components through a learnable visual prompt repository and prompt prefix tuning,ensuring that the relationships between triggers and objects are explicitly captured.To separate trigger and object representations in the visual prompt repository,we introduce the trigger–object separation and diversity losses that aids in disentangling trigger and object visual features.Next,by aligning image features with feature decomposition and fusion,as well as learned contextual prompt tokens in a shared multimodal space,DBOM enables zero-shot generalization to novel trigger-object pairings that were unseen during training,thereby offering deeper insights into adversarial attack patterns.Experimental results on CIFAR-10 and GTSRB demonstrate that DBOM robustly detects poisoned images prior to downstream training,significantly enhancing the security of DNN training pipelines.
基金Supported by the National Natural Science Foundation of China(61210007).
文摘Learning disentangled representation of data is a key problem in deep learning.Specifically,disentangling 2D facial landmarks into different factors(e.g.,identity and expression)is widely used in the applications of face reconstruction,face reenactment and talking head et al..However,due to the sparsity of landmarks and the lack of accurate labels for the factors,it is hard to learn the disentangled representation of landmarks.To address these problem,we propose a simple and effective model named FLD-VAE to disentangle arbitrary facial landmarks into identity and expression latent representations,which is based on a Variational Autoencoder framework.Besides,we propose three invariant loss functions in both latent and data levels to constrain the invariance of representations during training stage.Moreover,we implement an identity preservation loss to further enhance the representation ability of identity factor.To the best of our knowledge,this is the first work to end-to-end disentangle identity and expression factors simultaneously from one single facial landmark.
基金sponsored in part by the National Key Research and Development Program of China(No.2023YFB3307500)the Science and Technology Innovation Project of Hunan Province(No.2023RC4014)the National Natural Science Foundation of China(NSFC)(Grant Nos.62076146,62021002,U20A6003,6212780016).
文摘1 Introduction Molecular Property Prediction aims to identify molecules sharing similar efficacious properties[1],which is a foundational task in drug discovery,materials science and bioinformatics.Graph neural networks(GNNs)have shown significant success in this field.However,GNN-based methods often face label scarcity,limiting their performance in predicting molecular properties.Besides,GNNs trained on specific datasets frequently struggle with generalization due to domain shift[2].
基金supported by the National Natural Science Foundation of China(No.U19A2059)the 2022 Research Foundation of Chengdu Textile College(No.X22032161).
文摘Currently,most existing inductive relation prediction approaches are based on subgraph structures,with subgraph features extracted using graph neural networks to predict relations.However,subgraphs may contain disconnected regions,which usually represent different semantic ranges.Because not all semantic information about the regions is helpful in relation prediction,we propose a relation prediction model based on a disentangled subgraph structure and implement a feature updating approach based on relevant semantic aggregation.To indirectly achieve the disentangled subgraph structure from a semantic perspective,the mapping of entity features into different semantic spaces and the aggregation of related semantics on each semantic space are updated.The disentangled model can focus on features having higher semantic relevance in the prediction,thus addressing a problem with existing approaches,which ignore the semantic differences in different subgraph structures.Furthermore,using a gated recurrent neural network,this model enhances the features of entities by sorting them by distance and extracting the path information in the subgraphs.Experimentally,it is shown that when there are numerous disconnected regions in the subgraph,our model outperforms existing mainstream models in terms of both Area Under the Curve-Precision-Recall(AUC-PR)and Hits@10.Experiments prove that semantic differences in the knowledge graph can be effectively distinguished and verify the effectiveness of this method.
文摘In recent years,the concept of digital human has attracted widespread attention from all walks of life,and the modelling of high-fidelity human bodies,heads,and hands has been intensively studied.This paper focuses on head modelling and proposes a generic head parametric model based on neural radiance fields.Specifically,we first use face recognition networks and 3D facial expression database FaceWarehouse to parameterize identity and expression semantics,respectively,and use both as conditional inputs to build a neural radiance field for the human head,thereby improving the head model’s representation ability while ensuring editing capabilities for the identity and expression of the rendered results;then,through a combination of volume rendering and neural rendering,the 3D representation of the head is rapidly rendered into the 2D plane,producing a high-fidelity image of the human head.Thanks to the well-designed loss functions and good implicit representation of the neural radiance field,our model can not only edit the identity and expression independently,but also freely modify the virtual camera position of the rendering results.It has excellent multi-view consistency,and has many applications in novel view synthesis,pose driving and more.
基金funded by the National Natural Science Foundation of China Project"Research on Intelligent Detection Techniques of Encrypted Malicious Traffic for Large-Scale Networks"(Grant No.62176264).
文摘Deep neural networks are known to be vulnerable to adversarial attacks.Unfortunately,the underlying mechanisms remain insufficiently understood,leading to empirical defenses that often fail against new attacks.In this paper,we explain adversarial attacks from the perspective of robust features,and propose a novel Generative Adversarial Network(GAN)-based Robust Feature Disentanglement framework(GRFD)for adversarial defense.The core of GRFD is an adversarial disentanglement structure comprising a generator and a discriminator.For the generator,we introduce a novel Latent Variable Constrained Variational Auto-Encoder(LVCVAE),which enhances the typical beta-VAE with a constrained rectification module to enforce explicit clustering of latent variables.To supervise the disentanglement of robust features,we design a Robust Supervisory Model(RSM)as the discriminator,sharing architectural alignment with the target model.The key innovation of RSM is our proposed Feature Robustness Metric(FRM),which serves as part of the training loss and synthesizes the classification ability of features as well as their resistance to perturbations.Extensive experiments on three benchmark datasets demonstrate the superiority of GRFD:it achieves 93.69%adversarial accuracy on MNIST,77.21%on CIFAR10,and 58.91%on CIFAR100 with minimal degradation in clean accuracy.Codes are available at:(accessed on 23 July 2025).
基金supported by the National Natural Science Foundation of China(82071500,82271540,32370724,82401759,81871055,32070679)Shanghai Clinical Research Center for Mental Health(19MC1911100)+11 种基金Shanghai Key Laboratory of Psychotic Disorders(13dz2260500)Shanghai Municipal Administrator of Traditional Chinese Medicine(ZY-(2021-2023)-0207-01)Shanghai Municipal Health Commission Collaborative Innovation Group(2024CXJQ03)Shanghai Science and Technology Innovation Action Program(24JS2840400,24ZR1439900,21Y11921100)Shanghai Municipal Science and Technology Major Project,the National Key R&D Program of China(2023YFA0913804,2024YFA0916603,2022FYC2503300)the Program of Shanghai Academic/Technology Research Leader(21XD1423300)Shanghai Pujiang Program(21PJD063)Shanghai Municipal Science and Technology Major Project(2017SHZDZX01)Shanghai Municipal Commission of Education(2024AIZD016)the National Key R&D Program of China(2019YFA0905400,2017YFC0908105,2021YFC2702100)National Program for Support of Top-Notch Young Professionals,Taishan Scholar Program of Shandong Province(tstp20240526)the Natural Science Foundation of Shandong Province(ZR2019YQ14,YDZX2021009,2021ZDSYS06).
文摘Background Numerous studies have consistently demonstrated that a considerable proportion of patients with major depressive disorder (MDD) frequently exhibit pronounced dyslipidaemia. However, the causal dynamics between MDD and dyslipidaemia remain elusive.Aims To comprehensively disentangle the genetic causality between MDD and various phenotypes of blood lipids, thereby facilitating the advancement of management strategies for these conditions.Methods We conducted a two-sample univariable Mendelian randomisation (MR) analysis using different models, including the inverse variance weighted (IVW) method and causal analysis using the summary effect (CAUSE) estimates, as well as a multivariable MR analysis. This analysis used summary statistics from genome-wide association studies (GWAS) of MDD and five lipid traits: low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, non-high-density lipoprotein cholesterol, total cholesterol and triglycerides (TG), encompassing 5 237 893 individuals of European and East Asian ancestries. For MDD, a total of 598 701 individuals were included, with 500 199 individuals of European ancestry (Ncase=170 756, Ncontrol=329 443) and 98 502 of East Asian ancestry (Ncase=12 588, Ncontrol=85 914). Lipid data were collected from 4 639 192 individuals through the Global Lipids Genetics Consortium (European, N=4 096 085;East Asian, N=543 107). Next, we used the two-step MR to explore the mediating factors between MDD and TG, and the risk factors affecting TG through MDD. Finally, we conducted a GWAS meta-analysis and enrichment analysis.Results In univariable MR, we observed a negative causal effect of low-density lipoprotein on MDD in both European populations (IVW: odds ratio (OR): 0.972, 95% confidence interval (CI) 0.947 to 0.998, p=0.037) and East Asian populations (IVW: OR: 0.928, 95% CI 0.864 to 0.997, p=0.042). Additionally, we identified a bidirectional causal relationship between TG and MDD, with TG having a causal effect on MDD (IVW: OR: 1.052, 95% CI 1.020 to 1.085, p=0.001) and MDD having a causal effect on TG (IVW: OR: 1.075, 95% CI 1.047 to 1.104, p<0.001). Multivariable MR analysis further supported the role of TG in MDD (OR: 1.205, 95% CI 1.034 to 1.405, p=0.017). CAUSE estimates indicated that the causal model of MDD on TG provided a better fit than the sharing model (p=0.003), while the association of TG on MDD was more likely due to horizontal correlated pleiotropy than causality. Mediation analyses revealed that waist-hip ratio (WHR) mediated 69% of the total causal effect of MDD on TG, while other identified risk factors exhibited lower mediating proportions either mediated through MDD (≤17%) or originating from MDD (≤29%). The GWAS meta-analysis highlighted potential pathways related to lipid processes and nucleosome assembling, with significant cell types identified in brain regions and liver tissues.Conclusions The findings indicate that genetic proxies of MDD are associated with elevated levels of TG, with WHR serving as a clinical indicator of the association. This suggests that interventions targeting WHR may be effective in reducing TG levels in patients with MDD.
基金supported by the National Natural Science Foundation of China(Nos.61825103,62202349)the Natural Science Foundation of Hubei Province(Nos.2022CFB352,2020CFA001)the Key Research&Development of Hubei Province(No.2020BIB006).
文摘representation that can identify and isolate different potential variables hidden in the highdimensional observations.Disentangled representation learning can capture information about a single change factor and control it by the corresponding potential subspace,providing a robust representation for complex changes in the data.In this paper,we first introduce and analyze the current status of research on disentangled representation and its causal mechanisms and summarize three crucial properties of disentangled representation.Then,disentangled representation learning algorithms are classified into four categories and outlined in terms of both mathematical description and applicability.Subsequently,the loss functions and objective evaluation metrics commonly used in existing work on disentangled representation are classified.Finally,the paper summarizes representative applications of disentangled representation learning in the field of remote sensing and discusses its future development.
基金supported by the National Key R&D Program of China(Nos.2020YFA0713601 and 2023YFA1008800)the National Natural Science Foundation of China(Nos.22341304,22341303,22103079 and 22073092)the Cooperation Project between Jilin Province and CAS(No.2023SYHZ0003).
文摘Using molecular dynamics(MD)simulations,this study explores the fluid properties of three polymer melts with the same number of entanglements,Z,achieved by adjusting the entanglement length Ne,while investigating the evolution of polymer melt conformation and entanglement under high-rate elongational flow.The identification of a master curve indicates consistent normalized linear viscoelastic behavior.Surprising findings regarding the steady-state viscosity at various elongational rates(Wi_(R)>4.7)for polymer melts with the same Z have been uncovered,challenging existing tube models.Nevertheless,the study demonstrates the potential for normalizing the steady-state elongational viscosity at high rates(Wi_(R)>4.7)by scaling with the square of the chain contour length.Additionally,the observed independence of viscosity on the elongational rate at high rates suggests that higher rates lead to a more significant alignment of polymer chains,a decrease in entanglement,and a stretching in contour length of polymer chains.Molecular-level tracking of tagged chains further supports the assumption of no entanglement under rapid elongation,emphasizing the need for further research on disentanglement in polymer melts subjected to high-rate elongational flow.These results carry significant implications for understanding and predicting the behavior of polymer melts under high-rate elongational flow conditions.
基金financially supported by the National Key R&D Program of China (2022YFA1503503 and 2022YFA1503504)the Natural Science Foundation of China (22038003,22178100,22178101 and 22008066)+1 种基金the Innovation Program of Shanghai Municipal Education Commission,the Program of Shanghai Academic/Technology Research Leader (21XD1421000)the Shanghai Science and Technology Innovation Action Plan (22JC1403800).
文摘Tailoring the electronic metal-support interaction(EMSI)has attracted considerable interests as one of the most efficient approaches to improve both the activity and stability of metal catalysts in heterogeneous catalysis.In this viewpoint,we illustrate the methodology and relevant fundamentals on the disentanglement,characterization,and interpretation of EMSI.Under the choice of monometallic catalyst over inert support,a combination of optimal experiment design,multimodal techniques,in situ characterization,with a comprehensive understanding of the underlying measurement protocols is highly desirable for a reliable determination of EMSI.Accordingly,not only the d-band filling but also d-band energy within the EMSI should be taken into consideration for providing general principles to guide the electron-promoting catalytic reaction.
基金Supported by Foundation of the Education Department of Liaoning Province under Grant No.20060160the Natural Science Foundation of Zhejiang Province under Grant No.Y6100098+1 种基金the National Natural Scinece Foundation of China under Grant No.11074062the funding support from Hangzhou Normal University
文摘By taking into account spatial degrees of freedom of atoms, we study the internal-state disentanglement dynamics of two atoms interacting with a vacuum multi-mode noise field. We show that the complete internal-state disentanglement of the two atoms, caused due to the atomic spontaneous emission can be achieved in a finite time.
文摘The bilinear generating function for products of two Laguerre 2D polynomials with different arguments is calculated. It corresponds to the formula of Mehler for the generating function of products of two Hermite polynomials. Furthermore, the generating function for mixed products of Laguerre 2D and Hermite 2D polynomials and for products of two Hermite 2D polynomials is calculated. A set of infinite sums over products of two Laguerre 2D polynomials as intermediate step to the generating function for products of Laguerre 2D polynomials is evaluated but these sums possess also proper importance for calculations with Laguerre polynomials. With the technique of operator disentanglement some operator identities are derived in an appendix. They allow calculating convolutions of Gaussian functions combined with polynomials in one- and two-dimensional case and are applied to evaluate the discussed generating functions.
基金Project supported by the National Natural Science Foundation of China(Grant No.11175113)
文摘We explore the theoretical possibility of extending the usual squeezed state to those produced by nonlinear singlemode squeezing operators. We derive the wave functions of exp[-（ig/2）（（1-X2）1/2P ＋ P（1-X2）1/2）]｜0 in the coordinate representation. A new operator＇s disentangling formula is derived as a by-product.
基金supported by the National Natural Science Foundation of China under Grant Nos.10475056 and 10775097
文摘We recommend a new convenient method for disentangling some exponential operators and derive a set of new operator identities. Especially, we derive the normal odering form of exp [fa^＋a ＋ ga^2＋ ＋ ka^2] without appealing to Lie algebra method. Application of these formulas in solving some dynamic Hamiltonian is presented.
基金supported by National Natural Science Foundation of China under Grant Nos.60578014 and 10775023
文摘The disentanglement evolution of bipartite spin-1/2 system coupled to a common surrounding XY chain in transverse fields at nonzero temperature is studied in this letter. The dynamical process of the entanglement is numerically and analytically investigated. We find that thermal effects can enhance disentanglement if the entangled initial state of the central spins does not in the decoherence free space. The critical phenomenon of quantum phase transitions reflected in the disentanglement can be washed out by the thermal effect eventually.
基金supported by the National Natural Science Foundation of China under Grant Nos.10775097 and 10874174
文摘In the preceding paper [Commun. Theor. Phys. 51 （2009） 321] we have recommended a convenient method for disentangling exponential operators in the form of exp{B ＋ C}, trying to find an operator A that satisfies [A, B] = C, and [A, [A, B]] = 0, then from the Baker-Hausdorff formula we have exp{B ＋C} ： exp（B ＋ [A, B]} = e^A e^B e^-A. After arranging e^Ae^B = e^B e^A e^W, the disentangling exp{B ＋ C} = e^B e^W is obtained. In this work we use this method to two-mode case, especially, derive the normal ordering form of exp[h（a^＋a ＋ b^＋b） ＋ ga^＋b^＋ ＋ kab] without appealing to Lie algebra method.
基金supported by National Natural Science Foundation of China under Grant Nos. 60678022 and 10704001the Specialized Research Fund for the Doctoral Program of Higher Education under Grant No. 20060357008+2 种基金Anhui Provincial Natural Science Foundation under Grant No. 070412060the Talent Foundation of Anhui UniversityAnhui Key Laboratory of Information Materials and Devices (Anhui University)
文摘We investigate the entanglement evolution of two qubits that are initially in Werner state under the classical phase noise. We discuss the influence of mixture degree on disentanglement. It is showed that the more mixed the state, the shorter is the time of disentanglement.
基金the Ministry of Education and Science of the Russian Federation(grant 14.756.31.0001).
文摘In this article,two new algorithms are presented that convert a given data tensor train into either a Tucker decomposition with orthogonal matrix factors or a multi-scale entanglement renormalization ansatz(MERA).The Tucker core tensor is never explicitly computed but stored as a tensor train instead,resulting in both computationally and storage efficient algorithms.Both the multilinear Tucker-ranks as well as the MERA-ranks are automatically determined by the algorithm for a given upper bound on the relative approximation error.In addition,an iterative algorithm with low computational complexity based on solving an orthogonal Procrustes problem is proposed for the first time to retrieve optimal rank-lowering disentangler tensors,which are a crucial component in the construction of a low-rank MERA.Numerical experiments demonstrate the effectiveness of the proposed algorithms together with the potential storage benefit of a low-rank MERA over a tensor train.
文摘A straightforward simple proof is given that dark energy is the natural conse-quence of a quantum disentanglement physical process. Thus while the ordinary energy density of the cosmos is equal to half that of Hardy’s quantum probability of Entanglement i.e. where , the density of cosmic dark energy is consequently one minus divided by two i.e. . This result is in full agreement with all the numerous previous theoretical predictions as well as being in remarkable agreement with the overwhelming majority of cosmic accurate measurements and observations.
