The usage of mobile-phone among children increased significantly. Children are in their growing phase and cells of their body are rapidly dividing, therefore propagation of electro-magnetic (EM) radiation occurs quick...The usage of mobile-phone among children increased significantly. Children are in their growing phase and cells of their body are rapidly dividing, therefore propagation of electro-magnetic (EM) radiation occurs quickly in children. The aim of the present study was to evaluate the extent of mobile-phone usage as well as its possible health effect. A total number of 455 (398 children and 57 adults, 396 urban and 59 rural) students of age group ranging from 10-29 years participated in this study. An “Information Gathering Chronological (IGC) model” was used for the collection and evaluation of information. The four major parameters, i.e. demographic and public uniqueness, mobile-phone consumption patterns, grievance of the “forgetfulness” symptom to the subjects and awareness about the safety measures were included to get the concise information from participants. We have observed that the prevalence of “forgetfulness” was 23.95% among mobile-phone users. The incidence of overall “forgetfulness” symptoms was 23.59%, 17.46%, 25.00% and 37.50% in low (LU), normal (NU), moderate (MU) and heavy (HU) mobile-phone users respectively. A trend for risk for “forgetfulness” was observed in HU as compared to LU in overall mobile-phone users. Three folds and nearly five folds increased risk for “forgetfulness” was found among HU as compared to LU in children (p ≤ 0.0210) and urban area mobile-phone users respectively. No significant difference for “forgetfulness” symptoms was found in other categories (i.e. adult and rural mobile-phone users). These results suggested that the incidences of “forgetfulness” among children from urban area mobile-phone users were significantly increased.展开更多
This study addresses the maneuver evasion problem for medium-to-long-range air-to-air missiles by proposing a KAN-λ-PPO-based evasion algorithm.The algorithm introduces Kolmogorov-Arnold Networks(KAN)to mitigate the ...This study addresses the maneuver evasion problem for medium-to-long-range air-to-air missiles by proposing a KAN-λ-PPO-based evasion algorithm.The algorithm introduces Kolmogorov-Arnold Networks(KAN)to mitigate the catastrophic forgetting issue of Multilayer Perceptrons(MLP)in continual learning,while incorporatingλ-return to resolve sparse reward challenges in evasion scenarios.First,we model the evasion problem withλ-return and present the KAN-λ-PPO algorithm.Subsequently,we establish game environments based on the segmented ballistic characteristics of medium and long range missiles.During training,a joint reward function is designed by combining the miss distance and positional advantages to train the agent.Experiments evaluate four dimensions:(1)Performance comparison between KAN and MLP in value function approximation;(2)Catastrophic forgetting mitigation of KAN-λ-PPO in dual-task scenarios;(3)Continual learning capabilities across multiple evasion scenarios;(4)Quantitative analysis of agent strategy evolution and positional advantages.Empirical results demonstrate that KAN improves value function approximation accuracy by an order of magnitude compared with traditional MLP architectures.In continual learning tasks,the KAN-λ-PPO scheme exhibits significant knowledge retention,achieving performance improvements of 32.7% and 8.6%over MLP baselines in Task1→2 and Task2→3 transitions,respectively.Furthermore,the learned maneuver strategies outperform High-G Barrel Rolls(HGB)and S-maneuver tactics in securing positional advantages while accomplishing evasion.展开更多
Continual learning fault diagnosis(CLFD)has gained growing interest in mechanical systems for its ability to accumulate and transfer knowledge in dynamic fault diagnosis scenarios.However,existing CLFD methods typical...Continual learning fault diagnosis(CLFD)has gained growing interest in mechanical systems for its ability to accumulate and transfer knowledge in dynamic fault diagnosis scenarios.However,existing CLFD methods typically assume balanced task distributions,neglecting the long-tailed nature of real-world fault occurrences,where certain faults dominate while others are rare.Due to the long-tailed distribution among different me-chanical conditions,excessive attention has been focused on the dominant type,leading to performance de-gradation in rarer types.In this paper,decoupling incremental classifier and representation learning(DICRL)is proposed to address the dual challenges of catastrophic forgetting introduced by incremental tasks and the bias in long-tailed CLFD(LT-CLFD).The core innovation lies in the structural decoupling of incremental classifier learning and representation learning.An instance-balanced sampling strategy is employed to learn more dis-criminative deep representations from the exemplars selected by the herding algorithm and new data.Then,the previous classifiers are frozen to prevent damage to representation learning during backward propagation.Cosine normalization classifier with learnable weight scaling is trained using a class-balanced sampling strategy to enhance classification accuracy.Experimental results demonstrate that DICRL outperforms existing continual learning methods across multiple benchmarks,demonstrating superior performance and robustness in both LT-CLFD and conventional CLFD.DICRL effectively tackles both catastrophic forgetting and long-tailed distribution in CLFD,enabling more reliable fault diagnosis in industrial applications.展开更多
Mental-health risk detection seeks early signs of distress from social media posts and clinical transcripts to enable timely intervention before crises.When such risks go undetected,consequences can escalate to self-h...Mental-health risk detection seeks early signs of distress from social media posts and clinical transcripts to enable timely intervention before crises.When such risks go undetected,consequences can escalate to self-harm,long-term disability,reduced productivity,and significant societal and economic burden.Despite recent advances,detecting risk from online text remains challenging due to heterogeneous language,evolving semantics,and the sequential emergence of new datasets.Effective solutions must encode clinically meaningful cues,reason about causal relations,and adapt to new domains without forgetting prior knowledge.To address these challenges,this paper presents a Continual Neuro-Symbolic Graph Learning(CNSGL)framework that unifies symbolic reasoning,causal inference,and continual learning within a single architecture.Each post is represented as a symbolic graph linking clinically relevant tags to textual content,enriched with causal edges derived from directional Point-wise Mutual Information(PMI).A two-layer Graph Convolutional Network(GCN)encodes these graphs,and a Transformer-based attention pooler aggregates node embeddings while providing interpretable tag-level importances.Continual adaptation across datasets is achieved through the Multi-Head Freeze(MH-Freeze)strategy,which freezes a shared encoder and incrementally trains lightweight task-specific heads(small classifiers attached to the shared embedding).Experimental evaluations across six diverse mental-health datasets ranging from Reddit discourse to clinical interviews,demonstrate that MH-Freeze consistently outperforms existing continual-learning baselines in both discriminative accuracy and calibration reliability.Across six datasets,MH-Freeze achieves up to 0.925 accuracy and 0.923 F1-Score,with AUPRC≥0.934 and AUROC≥0.942,consistently surpassing all continual-learning baselines.The results confirm the framework’s ability to preserve prior knowledge,adapt to domain shifts,and maintain causal interpretability,establishing CNSGL as a promising step toward robust,explainable,and lifelong mental-health risk assessment.展开更多
Modern intelligent systems,such as autonomous vehicles and face recognition,must continuously adapt to new scenarios while preserving their ability to handle previously encountered situations.However,when neural netwo...Modern intelligent systems,such as autonomous vehicles and face recognition,must continuously adapt to new scenarios while preserving their ability to handle previously encountered situations.However,when neural networks learn new classes sequentially,they suffer from catastrophic forgetting—the tendency to lose knowledge of earlier classes.This challenge,which lies at the core of class-incremental learning,severely limits the deployment of continual learning systems in real-world applications with streaming data.Existing approaches,including rehearsalbased methods and knowledge distillation techniques,have attempted to address this issue but often struggle to effectively preserve decision boundaries and discriminative features under limited memory constraints.To overcome these limitations,we propose a support vector-guided framework for class-incremental learning.The framework integrates an enhanced feature extractor with a Support Vector Machine classifier,which generates boundary-critical support vectors to guide both replay and distillation.Building on this architecture,we design a joint feature retention strategy that combines boundary proximity with feature diversity,and a Support Vector Distillation Loss that enforces dual alignment in decision and semantic spaces.In addition,triple attention modules are incorporated into the feature extractor to enhance representation power.Extensive experiments on CIFAR-100 and Tiny-ImageNet demonstrate effective improvements.On CIFAR-100 and Tiny-ImageNet with 5 tasks,our method achieves 71.68%and 58.61%average accuracy,outperforming strong baselines by 3.34%and 2.05%.These advantages are consistently observed across different task splits,highlighting the robustness and generalization of the proposed approach.Beyond benchmark evaluations,the framework also shows potential in few-shot and resource-constrained applications such as edge computing and mobile robotics.展开更多
Class-incremental learning studies the problem of continually learning new classes from data streams.But networks suffer from catastrophic forgetting problems,forgetting past knowledge when acquiring new knowledge.Amo...Class-incremental learning studies the problem of continually learning new classes from data streams.But networks suffer from catastrophic forgetting problems,forgetting past knowledge when acquiring new knowledge.Among different approaches,replay methods have shown exceptional promise for this challenge.But performance still baffles from two aspects:(i)data in imbalanced distribution and(ii)networks with semantic inconsistency.First,due to limited memory buffer,there exists imbalance between old and new classes.Direct optimisation would lead feature space skewed towards new classes,resulting in performance degradation on old classes.Second,existing methods normally leverage previous network to regularise the present network.However,the previous network is not trained on new classes,which means that these two networks are semantic inconsistent,leading to misleading guidance information.To address these two problems,we propose BCSD(BiaMix contrastive learning and memory similarity distillation).For imbalanced distribution,we design Biased MixUp,where mixed samples are in high weight from old classes and low weight from new classes.Thus,network learns to push decision boundaries towards new classes.We further leverage label information to construct contrastive learning in order to ensure discriminability.Meanwhile,for semantic inconsistency,we distill knowledge from the previous network by capturing the similarity of new classes in current tasks to old classes from the memory buffer and transfer that knowledge to the present network.Empirical results on various datasets demonstrate its effectiveness and efficiency.展开更多
President Xi Jinping’s New Year message calls for creating a better future for the world where every effort counts.IN his New Year message,President Xi Jinping called 2024 an extraordinary year with unforgettable mom...President Xi Jinping’s New Year message calls for creating a better future for the world where every effort counts.IN his New Year message,President Xi Jinping called 2024 an extraordinary year with unforgettable moments.China saw rainbows despite“winds and rains.”展开更多
Our research reveals the critical role of the suprachiasmatic nucleus(SCN)vasoactive intestinal peptide(VIP)neurons in mediating light-induced transient forgetting.Acute exposure to bright light selectively impairs tr...Our research reveals the critical role of the suprachiasmatic nucleus(SCN)vasoactive intestinal peptide(VIP)neurons in mediating light-induced transient forgetting.Acute exposure to bright light selectively impairs trace fear memory by activating VIP neurons in the SCN,as demonstrated by increased c-Fos expression and Ca2+recording.This effect can be replicated and reversed through optogenetic and chemogenetic manipulations of SCN VIP neurons.Furthermore,we identify the SCN→PVT(paraventricular nucleus of the thalamus)VIP neuronal circuitry as essential in this process.These findings establish a novel role for SCN VIP neurons in modulating memory accessibility in response to environmental light cues,extending their known function beyond circadian regulation and revealing a mechanism for transient forgetting.展开更多
The field of biometric identification has seen significant advancements over the years,with research focusing on enhancing the accuracy and security of these systems.One of the key developments is the integration of d...The field of biometric identification has seen significant advancements over the years,with research focusing on enhancing the accuracy and security of these systems.One of the key developments is the integration of deep learning techniques in biometric systems.However,despite these advancements,certain challenges persist.One of the most significant challenges is scalability over growing complexity.Traditional methods either require maintaining and securing a growing database,introducing serious security challenges,or relying on retraining the entiremodelwhen new data is introduced-a process that can be computationally expensive and complex.This challenge underscores the need for more efficient methods to scale securely.To this end,we introduce a novel approach that addresses these challenges by integrating multimodal biometrics,cancelable biometrics,and incremental learning techniques.This work is among the first attempts to seamlessly incorporate deep cancelable biometrics with dynamic architectural updates,applied incrementally to the deep learning model as new users are enrolled,achieving high performance with minimal catastrophic forgetting.By leveraging a One-Dimensional Convolutional Neural Network(1D-CNN)architecture combined with a hybrid incremental learning approach,our system achieves high recognition accuracy,averaging 98.98% over incrementing datasets,while ensuring user privacy through cancelable templates generated via a pre-trained CNN model and random projection.The approach demonstrates remarkable adaptability,utilizing the least intrusive biometric traits like facial features and fingerprints,ensuring not only robust performance but also long-term serviceability.展开更多
Vaccination is critical for controlling infectious diseases,but negative vaccination information can lead to vaccine hesitancy.To study how the interplay between information diffusion and disease transmission impacts ...Vaccination is critical for controlling infectious diseases,but negative vaccination information can lead to vaccine hesitancy.To study how the interplay between information diffusion and disease transmission impacts vaccination and epidemic spread,we propose a novel two-layer multiplex network model that integrates an unaware-acceptant-negative-unaware(UANU)information diffusion model with a susceptible-vaccinated-exposed-infected-susceptible(SVEIS)epidemiological framework.This model includes individual exposure and vaccination statuses,time-varying forgetting probabilities,and information conversion thresholds.Through the microscopic Markov chain approach(MMCA),we derive dynamic transition equations and the epidemic threshold expression,validated by Monte Carlo simulations.Using MMCA equations,we predict vaccination densities and analyze parameter effects on vaccination,disease transmission,and the epidemic threshold.Our findings suggest that promoting positive information,curbing the spread of negative information,enhancing vaccine effectiveness,and promptly identifying asymptomatic carriers can significantly increase vaccination rates,reduce epidemic spread,and raise the epidemic threshold.展开更多
Cardiovascular diseases(CVDs)continue to present a leading cause ofmortalityworldwide,emphasizing the importance of early and accurate prediction.Electrocardiogram(ECG)signals,central to cardiac monitoring,have increa...Cardiovascular diseases(CVDs)continue to present a leading cause ofmortalityworldwide,emphasizing the importance of early and accurate prediction.Electrocardiogram(ECG)signals,central to cardiac monitoring,have increasingly been integratedwithDeep Learning(DL)for real-time prediction of CVDs.However,DL models are prone to performance degradation due to concept drift and to catastrophic forgetting.To address this issue,we propose a realtime CVDs prediction approach,referred to as ADWIN-GFR that combines Convolutional Neural Network(CNN)layers,for spatial feature extraction,with Gated Recurrent Units(GRU),for temporal modeling,alongside adaptive drift detection and mitigation mechanisms.The proposed approach integratesAdaptiveWindowing(ADWIN)for realtime concept drift detection,a fine-tuning strategy based on Generative Features Replay(GFR)to preserve previously acquired knowledge,and a dynamic replay buffer ensuring variance,diversity,and data distribution coverage.Extensive experiments conducted on the MIT-BIH arrhythmia dataset demonstrate that ADWIN-GFR outperforms standard fine-tuning techniques,achieving an average post-drift accuracy of 95.4%,amacro F1-score of 93.9%,and a remarkably low forgetting score of 0.9%.It also exhibits an average drift detection delay of 12 steps and achieves an adaptation gain of 17.2%.These findings underscore the potential of ADWIN-GFR for deployment in real-world cardiac monitoring systems,including wearable ECG devices and hospital-based patient monitoring platforms.展开更多
This paper presents a method for estimating the parameters of DC-link capacitors in three-level NPC voltage source inverters(3L-NPC-VSI)used in grid-tied systems.The technique uses the signals generated by the intermo...This paper presents a method for estimating the parameters of DC-link capacitors in three-level NPC voltage source inverters(3L-NPC-VSI)used in grid-tied systems.The technique uses the signals generated by the intermodulation caused by the PWM strategy and converter topology interaction to estimate the capacitor parameters of the converter DC-link.It utilizes an observer-based structure consisting of a recursive noninteger sliding discrete Fourier transform(rnSDFT)and an RLS filter improved with a forgetting factor(oSDFT-RLS)to accurately estimate the capacitance and equivalent series resistance(ESR).Importantly,this method does not require additional sensors beyond those already installed in off-the-shelf 3L-NPC-VSI systems,ensuring its noninvasiveness.Furthermore,the oSDFTRLS estimates capacitor parameters in the time-frequency domain,enabling the tracking of capacitor degradation and predicting potential faults.Experimental results from the laboratory setup demonstrate the effectiveness of the proposed condition monitoring method.展开更多
A polynomial model, time origin shifting model(TOSM, is used to describe the trajectory of a moving target .Based on TOSM, a recursive laeast squares(RLS) algorithm with varied forgetting factor is derived for tracki...A polynomial model, time origin shifting model(TOSM, is used to describe the trajectory of a moving target .Based on TOSM, a recursive laeast squares(RLS) algorithm with varied forgetting factor is derived for tracking of a non-maneuvering target. In order to apply this algorithm to maneuvering targets tracking ,a tracking signal is performed on-line to determine what kind of TOSm will be in effect to track a target with different dynamics. An effective multiple model least squares filtering and forecasting method dadpted to real tracking of a maneuvering target is formulated. The algorithm is computationally more effcient than Kalman filter and the percentage improvement from simulations show both of them are considerably alike to some extent.展开更多
In this paper, we discuss questions of creating an electronic intellectual analogue of a human being. We introduce a mathematical concept of stimulus generating emotions. We also introduce a definition of logical thin...In this paper, we discuss questions of creating an electronic intellectual analogue of a human being. We introduce a mathematical concept of stimulus generating emotions. We also introduce a definition of logical thinking of robots and a notion of efficiency coefficient to describe their efficiency of rote (mechanical) memorizing. The paper proves theorems describing properties of permanent conflicts between logical and emotional thinking of robots with a nonabsolute rote memory.展开更多
有一句英文"Tell me,and I will forget;show me,and I may remember;involve me,and I can learn"(直译为:告诉我,我会忘记;展示给我看,我也许能记得;让我参与其中,我才能学会)。无论它的来源是谁,这句话阐述了一个简单而深...有一句英文"Tell me,and I will forget;show me,and I may remember;involve me,and I can learn"(直译为:告诉我,我会忘记;展示给我看,我也许能记得;让我参与其中,我才能学会)。无论它的来源是谁,这句话阐述了一个简单而深刻的道理:老师进行完整而高效的知识讲授,和学生真正掌握了这些知识之间,展开更多
In this paper the integrals of entwining structure (A,C,ψ) are discussed, where A is a k-algebra, C a k-coalgebra and a k-linear map. We prove that there exists a normalized integral γ:C→Hom(C,A) of (A,C,ψ) if and...In this paper the integrals of entwining structure (A,C,ψ) are discussed, where A is a k-algebra, C a k-coalgebra and a k-linear map. We prove that there exists a normalized integral γ:C→Hom(C,A) of (A,C,ψ) if and only if any representation of (A,C,ψ) is injective in a functorial way as a corepresentation of C. We give the dual results as well.展开更多
文摘The usage of mobile-phone among children increased significantly. Children are in their growing phase and cells of their body are rapidly dividing, therefore propagation of electro-magnetic (EM) radiation occurs quickly in children. The aim of the present study was to evaluate the extent of mobile-phone usage as well as its possible health effect. A total number of 455 (398 children and 57 adults, 396 urban and 59 rural) students of age group ranging from 10-29 years participated in this study. An “Information Gathering Chronological (IGC) model” was used for the collection and evaluation of information. The four major parameters, i.e. demographic and public uniqueness, mobile-phone consumption patterns, grievance of the “forgetfulness” symptom to the subjects and awareness about the safety measures were included to get the concise information from participants. We have observed that the prevalence of “forgetfulness” was 23.95% among mobile-phone users. The incidence of overall “forgetfulness” symptoms was 23.59%, 17.46%, 25.00% and 37.50% in low (LU), normal (NU), moderate (MU) and heavy (HU) mobile-phone users respectively. A trend for risk for “forgetfulness” was observed in HU as compared to LU in overall mobile-phone users. Three folds and nearly five folds increased risk for “forgetfulness” was found among HU as compared to LU in children (p ≤ 0.0210) and urban area mobile-phone users respectively. No significant difference for “forgetfulness” symptoms was found in other categories (i.e. adult and rural mobile-phone users). These results suggested that the incidences of “forgetfulness” among children from urban area mobile-phone users were significantly increased.
文摘This study addresses the maneuver evasion problem for medium-to-long-range air-to-air missiles by proposing a KAN-λ-PPO-based evasion algorithm.The algorithm introduces Kolmogorov-Arnold Networks(KAN)to mitigate the catastrophic forgetting issue of Multilayer Perceptrons(MLP)in continual learning,while incorporatingλ-return to resolve sparse reward challenges in evasion scenarios.First,we model the evasion problem withλ-return and present the KAN-λ-PPO algorithm.Subsequently,we establish game environments based on the segmented ballistic characteristics of medium and long range missiles.During training,a joint reward function is designed by combining the miss distance and positional advantages to train the agent.Experiments evaluate four dimensions:(1)Performance comparison between KAN and MLP in value function approximation;(2)Catastrophic forgetting mitigation of KAN-λ-PPO in dual-task scenarios;(3)Continual learning capabilities across multiple evasion scenarios;(4)Quantitative analysis of agent strategy evolution and positional advantages.Empirical results demonstrate that KAN improves value function approximation accuracy by an order of magnitude compared with traditional MLP architectures.In continual learning tasks,the KAN-λ-PPO scheme exhibits significant knowledge retention,achieving performance improvements of 32.7% and 8.6%over MLP baselines in Task1→2 and Task2→3 transitions,respectively.Furthermore,the learned maneuver strategies outperform High-G Barrel Rolls(HGB)and S-maneuver tactics in securing positional advantages while accomplishing evasion.
基金Supported by National Natural Science Foundation of China(Grant No.52272440)Suzhou Science Foundation(Grant Nos.SYG202323,ZXL2022027).
文摘Continual learning fault diagnosis(CLFD)has gained growing interest in mechanical systems for its ability to accumulate and transfer knowledge in dynamic fault diagnosis scenarios.However,existing CLFD methods typically assume balanced task distributions,neglecting the long-tailed nature of real-world fault occurrences,where certain faults dominate while others are rare.Due to the long-tailed distribution among different me-chanical conditions,excessive attention has been focused on the dominant type,leading to performance de-gradation in rarer types.In this paper,decoupling incremental classifier and representation learning(DICRL)is proposed to address the dual challenges of catastrophic forgetting introduced by incremental tasks and the bias in long-tailed CLFD(LT-CLFD).The core innovation lies in the structural decoupling of incremental classifier learning and representation learning.An instance-balanced sampling strategy is employed to learn more dis-criminative deep representations from the exemplars selected by the herding algorithm and new data.Then,the previous classifiers are frozen to prevent damage to representation learning during backward propagation.Cosine normalization classifier with learnable weight scaling is trained using a class-balanced sampling strategy to enhance classification accuracy.Experimental results demonstrate that DICRL outperforms existing continual learning methods across multiple benchmarks,demonstrating superior performance and robustness in both LT-CLFD and conventional CLFD.DICRL effectively tackles both catastrophic forgetting and long-tailed distribution in CLFD,enabling more reliable fault diagnosis in industrial applications.
基金supported by the National Research Foundation of Korea(NRF)grant funded by the Korea government(MSIT)(RS-2025-00518960)in part by the National Research Foundation of Korea(NRF)grant funded by the Korea government(MSIT)(RS-2025-00563192).
文摘Mental-health risk detection seeks early signs of distress from social media posts and clinical transcripts to enable timely intervention before crises.When such risks go undetected,consequences can escalate to self-harm,long-term disability,reduced productivity,and significant societal and economic burden.Despite recent advances,detecting risk from online text remains challenging due to heterogeneous language,evolving semantics,and the sequential emergence of new datasets.Effective solutions must encode clinically meaningful cues,reason about causal relations,and adapt to new domains without forgetting prior knowledge.To address these challenges,this paper presents a Continual Neuro-Symbolic Graph Learning(CNSGL)framework that unifies symbolic reasoning,causal inference,and continual learning within a single architecture.Each post is represented as a symbolic graph linking clinically relevant tags to textual content,enriched with causal edges derived from directional Point-wise Mutual Information(PMI).A two-layer Graph Convolutional Network(GCN)encodes these graphs,and a Transformer-based attention pooler aggregates node embeddings while providing interpretable tag-level importances.Continual adaptation across datasets is achieved through the Multi-Head Freeze(MH-Freeze)strategy,which freezes a shared encoder and incrementally trains lightweight task-specific heads(small classifiers attached to the shared embedding).Experimental evaluations across six diverse mental-health datasets ranging from Reddit discourse to clinical interviews,demonstrate that MH-Freeze consistently outperforms existing continual-learning baselines in both discriminative accuracy and calibration reliability.Across six datasets,MH-Freeze achieves up to 0.925 accuracy and 0.923 F1-Score,with AUPRC≥0.934 and AUROC≥0.942,consistently surpassing all continual-learning baselines.The results confirm the framework’s ability to preserve prior knowledge,adapt to domain shifts,and maintain causal interpretability,establishing CNSGL as a promising step toward robust,explainable,and lifelong mental-health risk assessment.
基金supported by the Gansu Provincial Natural Science Foundation(grant number 25JRRA074)the Gansu Provincial Key R&D Science and Technology Program(grant number 24YFGA060)the National Natural Science Foundation of China(grant number 62161019).
文摘Modern intelligent systems,such as autonomous vehicles and face recognition,must continuously adapt to new scenarios while preserving their ability to handle previously encountered situations.However,when neural networks learn new classes sequentially,they suffer from catastrophic forgetting—the tendency to lose knowledge of earlier classes.This challenge,which lies at the core of class-incremental learning,severely limits the deployment of continual learning systems in real-world applications with streaming data.Existing approaches,including rehearsalbased methods and knowledge distillation techniques,have attempted to address this issue but often struggle to effectively preserve decision boundaries and discriminative features under limited memory constraints.To overcome these limitations,we propose a support vector-guided framework for class-incremental learning.The framework integrates an enhanced feature extractor with a Support Vector Machine classifier,which generates boundary-critical support vectors to guide both replay and distillation.Building on this architecture,we design a joint feature retention strategy that combines boundary proximity with feature diversity,and a Support Vector Distillation Loss that enforces dual alignment in decision and semantic spaces.In addition,triple attention modules are incorporated into the feature extractor to enhance representation power.Extensive experiments on CIFAR-100 and Tiny-ImageNet demonstrate effective improvements.On CIFAR-100 and Tiny-ImageNet with 5 tasks,our method achieves 71.68%and 58.61%average accuracy,outperforming strong baselines by 3.34%and 2.05%.These advantages are consistently observed across different task splits,highlighting the robustness and generalization of the proposed approach.Beyond benchmark evaluations,the framework also shows potential in few-shot and resource-constrained applications such as edge computing and mobile robotics.
基金supported by the National Natural Science Foundation of China(62176188).
文摘Class-incremental learning studies the problem of continually learning new classes from data streams.But networks suffer from catastrophic forgetting problems,forgetting past knowledge when acquiring new knowledge.Among different approaches,replay methods have shown exceptional promise for this challenge.But performance still baffles from two aspects:(i)data in imbalanced distribution and(ii)networks with semantic inconsistency.First,due to limited memory buffer,there exists imbalance between old and new classes.Direct optimisation would lead feature space skewed towards new classes,resulting in performance degradation on old classes.Second,existing methods normally leverage previous network to regularise the present network.However,the previous network is not trained on new classes,which means that these two networks are semantic inconsistent,leading to misleading guidance information.To address these two problems,we propose BCSD(BiaMix contrastive learning and memory similarity distillation).For imbalanced distribution,we design Biased MixUp,where mixed samples are in high weight from old classes and low weight from new classes.Thus,network learns to push decision boundaries towards new classes.We further leverage label information to construct contrastive learning in order to ensure discriminability.Meanwhile,for semantic inconsistency,we distill knowledge from the previous network by capturing the similarity of new classes in current tasks to old classes from the memory buffer and transfer that knowledge to the present network.Empirical results on various datasets demonstrate its effectiveness and efficiency.
文摘President Xi Jinping’s New Year message calls for creating a better future for the world where every effort counts.IN his New Year message,President Xi Jinping called 2024 an extraordinary year with unforgettable moments.China saw rainbows despite“winds and rains.”
基金supported by grants from the National Natural Science Foundation of China(32021002 and 31900724)the STI2030-Major Projects(2022ZD0204900)the Tsinghua-Peking Joint Center for Life Sciences.
文摘Our research reveals the critical role of the suprachiasmatic nucleus(SCN)vasoactive intestinal peptide(VIP)neurons in mediating light-induced transient forgetting.Acute exposure to bright light selectively impairs trace fear memory by activating VIP neurons in the SCN,as demonstrated by increased c-Fos expression and Ca2+recording.This effect can be replicated and reversed through optogenetic and chemogenetic manipulations of SCN VIP neurons.Furthermore,we identify the SCN→PVT(paraventricular nucleus of the thalamus)VIP neuronal circuitry as essential in this process.These findings establish a novel role for SCN VIP neurons in modulating memory accessibility in response to environmental light cues,extending their known function beyond circadian regulation and revealing a mechanism for transient forgetting.
基金the Deputyship for Research&Innovation,Ministry of Education in Saudi Arabia for funding this research work through project number RI-44-0833.
文摘The field of biometric identification has seen significant advancements over the years,with research focusing on enhancing the accuracy and security of these systems.One of the key developments is the integration of deep learning techniques in biometric systems.However,despite these advancements,certain challenges persist.One of the most significant challenges is scalability over growing complexity.Traditional methods either require maintaining and securing a growing database,introducing serious security challenges,or relying on retraining the entiremodelwhen new data is introduced-a process that can be computationally expensive and complex.This challenge underscores the need for more efficient methods to scale securely.To this end,we introduce a novel approach that addresses these challenges by integrating multimodal biometrics,cancelable biometrics,and incremental learning techniques.This work is among the first attempts to seamlessly incorporate deep cancelable biometrics with dynamic architectural updates,applied incrementally to the deep learning model as new users are enrolled,achieving high performance with minimal catastrophic forgetting.By leveraging a One-Dimensional Convolutional Neural Network(1D-CNN)architecture combined with a hybrid incremental learning approach,our system achieves high recognition accuracy,averaging 98.98% over incrementing datasets,while ensuring user privacy through cancelable templates generated via a pre-trained CNN model and random projection.The approach demonstrates remarkable adaptability,utilizing the least intrusive biometric traits like facial features and fingerprints,ensuring not only robust performance but also long-term serviceability.
基金supported by the National Social Science Foundation of China(Grant Nos.21BGL217 and 22CGL050)the Philosophy and Social Science Fund of Education Department of Jiangsu Province(Grant No.2020SJA2346).
文摘Vaccination is critical for controlling infectious diseases,but negative vaccination information can lead to vaccine hesitancy.To study how the interplay between information diffusion and disease transmission impacts vaccination and epidemic spread,we propose a novel two-layer multiplex network model that integrates an unaware-acceptant-negative-unaware(UANU)information diffusion model with a susceptible-vaccinated-exposed-infected-susceptible(SVEIS)epidemiological framework.This model includes individual exposure and vaccination statuses,time-varying forgetting probabilities,and information conversion thresholds.Through the microscopic Markov chain approach(MMCA),we derive dynamic transition equations and the epidemic threshold expression,validated by Monte Carlo simulations.Using MMCA equations,we predict vaccination densities and analyze parameter effects on vaccination,disease transmission,and the epidemic threshold.Our findings suggest that promoting positive information,curbing the spread of negative information,enhancing vaccine effectiveness,and promptly identifying asymptomatic carriers can significantly increase vaccination rates,reduce epidemic spread,and raise the epidemic threshold.
基金supported by Princess Nourah bint Abdulrahman University Researchers Supporting Project number(PNURSP2025R196)Princess Nourah bint Abdulrahman University,Riyadh,Saudi Arabia.
文摘Cardiovascular diseases(CVDs)continue to present a leading cause ofmortalityworldwide,emphasizing the importance of early and accurate prediction.Electrocardiogram(ECG)signals,central to cardiac monitoring,have increasingly been integratedwithDeep Learning(DL)for real-time prediction of CVDs.However,DL models are prone to performance degradation due to concept drift and to catastrophic forgetting.To address this issue,we propose a realtime CVDs prediction approach,referred to as ADWIN-GFR that combines Convolutional Neural Network(CNN)layers,for spatial feature extraction,with Gated Recurrent Units(GRU),for temporal modeling,alongside adaptive drift detection and mitigation mechanisms.The proposed approach integratesAdaptiveWindowing(ADWIN)for realtime concept drift detection,a fine-tuning strategy based on Generative Features Replay(GFR)to preserve previously acquired knowledge,and a dynamic replay buffer ensuring variance,diversity,and data distribution coverage.Extensive experiments conducted on the MIT-BIH arrhythmia dataset demonstrate that ADWIN-GFR outperforms standard fine-tuning techniques,achieving an average post-drift accuracy of 95.4%,amacro F1-score of 93.9%,and a remarkably low forgetting score of 0.9%.It also exhibits an average drift detection delay of 12 steps and achieves an adaptation gain of 17.2%.These findings underscore the potential of ADWIN-GFR for deployment in real-world cardiac monitoring systems,including wearable ECG devices and hospital-based patient monitoring platforms.
基金funded by the Brazilian National Council for Scientific and Technological Development—CNPq(CNPq grant number 405997/2022-1)supported by the EMBRAPII VIRTUS Competence Center in Intelligent Hardware for Industry—VIRTUS-CC(MCTI grant number 055/2023).
文摘This paper presents a method for estimating the parameters of DC-link capacitors in three-level NPC voltage source inverters(3L-NPC-VSI)used in grid-tied systems.The technique uses the signals generated by the intermodulation caused by the PWM strategy and converter topology interaction to estimate the capacitor parameters of the converter DC-link.It utilizes an observer-based structure consisting of a recursive noninteger sliding discrete Fourier transform(rnSDFT)and an RLS filter improved with a forgetting factor(oSDFT-RLS)to accurately estimate the capacitance and equivalent series resistance(ESR).Importantly,this method does not require additional sensors beyond those already installed in off-the-shelf 3L-NPC-VSI systems,ensuring its noninvasiveness.Furthermore,the oSDFTRLS estimates capacitor parameters in the time-frequency domain,enabling the tracking of capacitor degradation and predicting potential faults.Experimental results from the laboratory setup demonstrate the effectiveness of the proposed condition monitoring method.
文摘A polynomial model, time origin shifting model(TOSM, is used to describe the trajectory of a moving target .Based on TOSM, a recursive laeast squares(RLS) algorithm with varied forgetting factor is derived for tracking of a non-maneuvering target. In order to apply this algorithm to maneuvering targets tracking ,a tracking signal is performed on-line to determine what kind of TOSm will be in effect to track a target with different dynamics. An effective multiple model least squares filtering and forecasting method dadpted to real tracking of a maneuvering target is formulated. The algorithm is computationally more effcient than Kalman filter and the percentage improvement from simulations show both of them are considerably alike to some extent.
文摘In this paper, we discuss questions of creating an electronic intellectual analogue of a human being. We introduce a mathematical concept of stimulus generating emotions. We also introduce a definition of logical thinking of robots and a notion of efficiency coefficient to describe their efficiency of rote (mechanical) memorizing. The paper proves theorems describing properties of permanent conflicts between logical and emotional thinking of robots with a nonabsolute rote memory.
文摘有一句英文"Tell me,and I will forget;show me,and I may remember;involve me,and I can learn"(直译为:告诉我,我会忘记;展示给我看,我也许能记得;让我参与其中,我才能学会)。无论它的来源是谁,这句话阐述了一个简单而深刻的道理:老师进行完整而高效的知识讲授,和学生真正掌握了这些知识之间,
文摘In this paper the integrals of entwining structure (A,C,ψ) are discussed, where A is a k-algebra, C a k-coalgebra and a k-linear map. We prove that there exists a normalized integral γ:C→Hom(C,A) of (A,C,ψ) if and only if any representation of (A,C,ψ) is injective in a functorial way as a corepresentation of C. We give the dual results as well.
