After billions of years of evolution,biological intelligence has converged on unrivalled energy efficiency and environmental adaptability.The human brain,for instance,is highly efficient in information transmission,co...After billions of years of evolution,biological intelligence has converged on unrivalled energy efficiency and environmental adaptability.The human brain,for instance,is highly efficient in information transmission,consuming only about 20 W onaverage in a resting state[1,2].A key to this efficiency is that biological signal transduction and processing rely significantly on multi-ions as the signal carriers.Inspired by this paradigm.展开更多
The goal of this paper is to investigate the long-time dynamics of solutions to a Kirchhoff type suspension bridge equation with nonlinear damping and memory term.For this problem we establish the well-posedness and e...The goal of this paper is to investigate the long-time dynamics of solutions to a Kirchhoff type suspension bridge equation with nonlinear damping and memory term.For this problem we establish the well-posedness and existence of uniform attractor under some suitable assumptions on the nonlinear term g(u),the nonlinear damping f(u_(t))and the external force h(x,t).Specifically,the asymptotic compactness of the semigroup is verified by the energy reconstruction method.展开更多
The highly dynamic nature,strong uncertainty,and coupled multiple safety constraints inherent in carrier aircraft recovery operations pose severe challenges for real-time decision-making.Addressing bolter scenarios,th...The highly dynamic nature,strong uncertainty,and coupled multiple safety constraints inherent in carrier aircraft recovery operations pose severe challenges for real-time decision-making.Addressing bolter scenarios,this study proposes an intelligent decision-making framework based on a deep long short-term memory Q-network.This framework transforms the real-time sequencing for bolter recovery problem into a partially observable Markov decision process.It employs a stacked long shortterm memory network to accurately capture the long-range temporal dependencies of bolter event chains and fuel consumption.Furthermore,it integrates a prioritized experience replay training mechanism to construct a safe and adaptive scheduling system capable of millisecond-level real-time decision-making.Experimental demonstrates that,within large-scale mass recovery scenarios,the framework achieves zero safety violations in static environments and maintains a fuel safety violation rate below 10%in dynamic scenarios,with single-step decision times at the millisecond level.The model exhibits strong generalization capability,effectively responding to unforeseen emergent situations—such as multiple bolters and fuel emergencies—without requiring retraining.This provides robust support for efficient carrier-based aircraft recovery operations.展开更多
Shape memory polymers used in 4D printing only had one permanent shape after molding,which limited their applications in requiring multiple reconstructions and multifunctional shapes.Furthermore,the inherent stability...Shape memory polymers used in 4D printing only had one permanent shape after molding,which limited their applications in requiring multiple reconstructions and multifunctional shapes.Furthermore,the inherent stability of the triazine ring structure within cyanate ester(CE)crosslinked networks after molding posed significant challenges for both recycling,repairing,and degradation of resin.To address these obstacles,dynamic thiocyanate ester(TCE)bonds and photocurable group were incorporated into CE,obtaining the recyclable and 3D printable CE covalent adaptable networks(CANs),denoted as PTCE1.5.This material exhibits a Young's modulus of 810 MPa and a tensile strength of 50.8 MPa.Notably,damaged printed PTCE1.5 objects can be readily repaired through reprinting and interface rejoining by thermal treatment.Leveraging the solid-state plasticity,PTCE1.5 also demonstrated attractive shape memory ability and permanent shape reconfigurability,enabling its reconfigurable 4D printing.The printed PTCE1.5 hinges and a main body were assembled into a deployable and retractable satellite model,validating its potential application as a controllable component in the aerospace field.Moreover,printed PTCE1.5 can be fully degraded into thiol-modified intermediate products.Overall,this material not only enriches the application range of CE resin,but also provides a reliable approach to addressing environmental issue.展开更多
As a typical in-memory computing hardware design, nonvolatile ternary content-addressable memories(TCAMs) enable the logic operation and data storage for high throughout in parallel big data processing. However,TCAM c...As a typical in-memory computing hardware design, nonvolatile ternary content-addressable memories(TCAMs) enable the logic operation and data storage for high throughout in parallel big data processing. However,TCAM cells based on conventional silicon-based devices suffer from structural complexity and large footprintlimitations. Here, we demonstrate an ultrafast nonvolatile TCAM cell based on the MoTe2/hBN/multilayergraphene (MLG) van der Waals heterostructure using a top-gated partial floating-gate field-effect transistor(PFGFET) architecture. Based on its ambipolar transport properties, the carrier type in the source/drain andcentral channel regions of the MoTe2 channel can be efficiently tuned by the control gate and top gate, respectively,enabling the reconfigurable operation of the device in either memory or FET mode. When working inthe memory mode, it achieves an ultrafast 60 ns programming/erase speed with a current on-off ratio of ∼105,excellent retention capability, and robust endurance. When serving as a reconfigurable transistor, unipolar p-typeand n-type FETs are obtained by adopting ultrafast 60 ns control-gate voltage pulses with different polarities.The monolithic integration of memory and logic within a single device enables the content-addressable memory(CAM) functionality. Finally, by integrating two PFGFETs in parallel, a TCAM cell with a high current ratioof ∼10^(5) between the match and mismatch states is achieved without requiring additional peripheral circuitry.These results provide a promising route for the design of high-performance TCAM devices for future in-memorycomputing applications.展开更多
Supporting real-time and privacy-preserving learning at the edge is emerging as a critical trend,bringing forth substantial challenges for deep learning(DL)training in the context of limited GPU(graphic processing uni...Supporting real-time and privacy-preserving learning at the edge is emerging as a critical trend,bringing forth substantial challenges for deep learning(DL)training in the context of limited GPU(graphic processing unit)memory.Recent work has sought to address the limitations by swapping tensors between GPU memory and CPU memory.Unfortunately,their tensor-based memory management encounters additional overhead since the swapped tensors do not align with the actual memory demands,resulting in decreased throughput.This paper introduces a vTensor-based memory management approach designed to mitigate memory swapping overhead.Virtualized tensors,dubbed vTensors,are used to finely align memory swapping amounts with real-time memory demands.Firstly,we introduce an abstraction layer that virtualizes coarse-grained tensors to multiple finer-grained vTensors.Secondly,we propose the Layered Graph Model(LGM)for analyzing vTensor mappings,which produces a memory swapping plan leveraged in the subsequent DL training iterations.Evaluations conducted on typical edge deep learning models illustrate that our approach surpasses prior work with a 15.60%increase in DL training throughput.展开更多
Working memory is a core cognitive function that supports goal-directed behavior and complex thought.We developed a spatial working memory and attention test on paired symbols(SWAPS)which has been proved to be a usefu...Working memory is a core cognitive function that supports goal-directed behavior and complex thought.We developed a spatial working memory and attention test on paired symbols(SWAPS)which has been proved to be a useful and valid tool for spatial working memory and attention studies in the fields of cognitive psychology,education,and psychiatry.The repeated administration of working memory capacity tests is common in clinical and research settings.Studies suggest that repeated cognitive tests may improve the performance scores also known as retest effects.The systematic investigation of retest effects in SWAPS is critical for interpreting scientific results,but it is still not fully developed.To address this,we recruited 77 college students aged 18–21 years and used SWAPS comprising 72 trials with different memory loads,learning time,and delay span.We repeated the test once a week for five weeks to investigate the retest effects of SWAPS.There were significant retest effects in the first two tests:the accuracy of the SWAPS tests significantly increased,and then stabilized.These findings provide useful information for researchers to appropriately use or interpret the repeated working memory tests.Further experiments are still needed to clarify the factors that mediate the retest effects,and find out the cognitive mechanism that influences the retest effects.展开更多
Memory deficit,which is often associated with aging and many psychiatric,neurological,and neurodegenerative diseases,has been a challenging issue for treatment.Up till now,all potential drug candidates have failed to ...Memory deficit,which is often associated with aging and many psychiatric,neurological,and neurodegenerative diseases,has been a challenging issue for treatment.Up till now,all potential drug candidates have failed to produce satisfa ctory effects.Therefore,in the search for a solution,we found that a treatment with the gene corresponding to the RGS14414protein in visual area V2,a brain area connected with brain circuits of the ventral stream and the medial temporal lobe,which is crucial for object recognition memory(ORM),can induce enhancement of ORM.In this study,we demonstrated that the same treatment with RGS14414in visual area V2,which is relatively unaffected in neurodegenerative diseases such as Alzheimer s disease,produced longlasting enhancement of ORM in young animals and prevent ORM deficits in rodent models of aging and Alzheimer’s disease.Furthermore,we found that the prevention of memory deficits was mediated through the upregulation of neuronal arbo rization and spine density,as well as an increase in brain-derived neurotrophic factor(BDNF).A knockdown of BDNF gene in RGS14414-treated aging rats and Alzheimer s disease model mice caused complete loss in the upregulation of neuronal structural plasticity and in the prevention of ORM deficits.These findings suggest that BDNF-mediated neuronal structural plasticity in area V2 is crucial in the prevention of memory deficits in RGS14414-treated rodent models of aging and Alzheimer’s disease.Therefore,our findings of RGS14414gene-mediated activation of neuronal circuits in visual area V2 have therapeutic relevance in the treatment of memory deficits.展开更多
The active development of space industry necessitates the cre-ation of novel materials with unique properties,including shape memory alloys(SMAs).The development of ultra-high temperature SMAs(UHTSMAs)with operating t...The active development of space industry necessitates the cre-ation of novel materials with unique properties,including shape memory alloys(SMAs).The development of ultra-high temperature SMAs(UHTSMAs)with operating temperatures above 400℃is a significant challenge[1-3].It is known that reversible thermoelas-tic martensitic transformation(MT)is the basis for shape mem-ory behavior[4].Currently,there are several systems in which MT temperatures meet the above requirements,for example,RuNb[5],HfPd[6],TiPd[7].展开更多
This study investigates the application of Learnable Memory Vision Transformers(LMViT)for detecting metal surface flaws,comparing their performance with traditional CNNs,specifically ResNet18 and ResNet50,as well as o...This study investigates the application of Learnable Memory Vision Transformers(LMViT)for detecting metal surface flaws,comparing their performance with traditional CNNs,specifically ResNet18 and ResNet50,as well as other transformer-based models including Token to Token ViT,ViT withoutmemory,and Parallel ViT.Leveraging awidely-used steel surface defect dataset,the research applies data augmentation and t-distributed stochastic neighbor embedding(t-SNE)to enhance feature extraction and understanding.These techniques mitigated overfitting,stabilized training,and improved generalization capabilities.The LMViT model achieved a test accuracy of 97.22%,significantly outperforming ResNet18(88.89%)and ResNet50(88.90%),aswell as the Token to TokenViT(88.46%),ViT without memory(87.18),and Parallel ViT(91.03%).Furthermore,LMViT exhibited superior training and validation performance,attaining a validation accuracy of 98.2%compared to 91.0%for ResNet 18,96.0%for ResNet50,and 89.12%,87.51%,and 91.21%for Token to Token ViT,ViT without memory,and Parallel ViT,respectively.The findings highlight the LMViT’s ability to capture long-range dependencies in images,an areawhere CNNs struggle due to their reliance on local receptive fields and hierarchical feature extraction.The additional transformer-based models also demonstrate improved performance in capturing complex features over CNNs,with LMViT excelling particularly at detecting subtle and complex defects,which is critical for maintaining product quality and operational efficiency in industrial applications.For instance,the LMViT model successfully identified fine scratches and minor surface irregularities that CNNs often misclassify.This study not only demonstrates LMViT’s potential for real-world defect detection but also underscores the promise of other transformer-based architectures like Token to Token ViT,ViT without memory,and Parallel ViT in industrial scenarios where complex spatial relationships are key.Future research may focus on enhancing LMViT’s computational efficiency for deployment in real-time quality control systems.展开更多
At the beginning of the 20^(th)century,German scientist Richard Semon introduced the term'engram'to describe the neural substrate implicated in the processes of memory formation and retrieval[1].The trace of t...At the beginning of the 20^(th)century,German scientist Richard Semon introduced the term'engram'to describe the neural substrate implicated in the processes of memory formation and retrieval[1].The trace of the corresponding biophysical and biochemical changes in the brain responding to an external stimulus is called an engram,and understanding the physical manifestations of memory formation and recall remains a fundamental yet unresolved question[2].展开更多
Dear Editor,Sleep and memory are highly linked across species.Sleep gates and stabilizes memory,critical for memory processing.Insufficient sleep impairs cognition acutely/chronically,in vertebrates and invertebrates[...Dear Editor,Sleep and memory are highly linked across species.Sleep gates and stabilizes memory,critical for memory processing.Insufficient sleep impairs cognition acutely/chronically,in vertebrates and invertebrates[1,2].While key elements are characterized[3,4],how a single molecule integrates sleep and memory remains unknown.展开更多
The gravitational memory effect manifests gravitational nonlinearity,degenerate vacua,and asymptotic symmetries;its detection is considered challenging.We propose using a space-borne interferometer to detect memory si...The gravitational memory effect manifests gravitational nonlinearity,degenerate vacua,and asymptotic symmetries;its detection is considered challenging.We propose using a space-borne interferometer to detect memory signals from stellar-mass binary black holes(BBHs),typically targeted by ground-based detectors.We use DECIGO detector as an example.Over 5 years,DECIGO is estimated to detect approximately 2,036 memory signals(SNRs>3)from stellar-mass BBHs.Simulations used frequency-domain memory waveforms for direct SNR estimation.Predictions utilized a GWTC-3 constrained BBH population model(Power law+Peak mass,DEFAULT spin,Madau-Dickinson merger rate).The analysis used conservative lower merger rate limits and considered orbital eccentricity.The high detection rate stems from strong memory signals within DECIGO’s bandwidth and the abundance of stellar-mass BBHs.This substantial and conservative detection count enables statistical use of the memory effect for fundamental physics and astrophysics.DECIGO exemplifies that space interferometers may better detect memory signals from smaller mass binaries than their typical targets.Detectors in lower frequency bands are expected to find strong memory signals from∼10^(4)M⊙binaries.展开更多
Shape memory alloys(SMAs)and shape memory ceramics(SMCs)exhibit high recovery ability due to the martensitic transformation,which complicates the fracture mechanism of SMAs and SMCs.The phase field method,as a powerfu...Shape memory alloys(SMAs)and shape memory ceramics(SMCs)exhibit high recovery ability due to the martensitic transformation,which complicates the fracture mechanism of SMAs and SMCs.The phase field method,as a powerful numerical simulation tool,can efficiently resolve the microstructural evolution,multi-field coupling effects,and fracture behavior of SMAs and SMCs.This review begins by presenting the fundamental theoretical framework of the fracture phase field method as applied to SMAs and SMCs,covering key aspects such as the phase field modeling of martensitic transformation and brittle fracture.Subsequently,it systematically examines the phase field simulations of fracture behaviors in SMAs and SMCs,with particular emphasis on how crystallographic orientation,grain size,and grain boundary properties influence the crack propagation.Additionally,the interplay between martensite transformation and fracture mechanisms is analyzed to provide deeper insights into the material responses under mechanical loading.Finally,the review explores future prospects and emerging trends in phase field simulations of SMA and SMC fracture behavior,along with potential advancements in the fracture phase field method itself,including multi-physics coupling and enhanced computational efficiency for large-scale simulations.展开更多
Computing-in-memory(CIM)has been a promising candidate for artificial-intelligent applications thanks to the absence of data transfer between computation and storage blocks.Resistive random access memory(RRAM)based CI...Computing-in-memory(CIM)has been a promising candidate for artificial-intelligent applications thanks to the absence of data transfer between computation and storage blocks.Resistive random access memory(RRAM)based CIM has the advantage of high computing density,non-volatility as well as high energy efficiency.However,previous CIM research has predominantly focused on realizing high energy efficiency and high area efficiency for inference,while little attention has been devoted to addressing the challenges of on-chip programming speed,power consumption,and accuracy.In this paper,a fabri-cated 28 nm 576K RRAM-based CIM macro featuring optimized on-chip programming schemes is proposed to address the issues mentioned above.Different strategies of mapping weights to RRAM arrays are compared,and a novel direct-current ADC design is designed for both programming and inference stages.Utilizing the optimized hybrid programming scheme,4.67×programming speed,0.15×power saving and 4.31×compact weight distribution are realized.Besides,this macro achieves a normalized area efficiency of 2.82 TOPS/mm2 and a normalized energy efficiency of 35.6 TOPS/W.展开更多
Hippocampal neuronal loss causes cognitive dysfunction in Alzheimer’s disease.Adult hippocampal neurogenesis is reduced in patients with Alzheimer’s disease.Exercise stimulates adult hippocampal neurogenesis in rode...Hippocampal neuronal loss causes cognitive dysfunction in Alzheimer’s disease.Adult hippocampal neurogenesis is reduced in patients with Alzheimer’s disease.Exercise stimulates adult hippocampal neurogenesis in rodents and improves memory and slows cognitive decline in patients with Alzheimer’s disease.However,the molecular pathways for exercise-induced adult hippocampal neurogenesis and improved cognition in Alzheimer’s disease are poorly understood.Recently,regulator of G protein signaling 6(RGS6)was identified as the mediator of voluntary running-induced adult hippocampal neurogenesis in mice.Here,we generated novel RGS6fl/fl;APP_(SWE) mice and used retroviral approaches to examine the impact of RGS6 deletion from dentate gyrus neuronal progenitor cells on voluntary running-induced adult hippocampal neurogenesis and cognition in an amyloid-based Alzheimer’s disease mouse model.We found that voluntary running in APP_(SWE) mice restored their hippocampal cognitive impairments to that of control mice.This cognitive rescue was abolished by RGS6 deletion in dentate gyrus neuronal progenitor cells,which also abolished running-mediated increases in adult hippocampal neurogenesis.Adult hippocampal neurogenesis was reduced in sedentary APP_(SWE) mice versus control mice,with basal adult hippocampal neurogenesis reduced by RGS6 deletion in dentate gyrus neural precursor cells.RGS6 was expressed in neurons within the dentate gyrus of patients with Alzheimer’s disease with significant loss of these RGS6-expressing neurons.Thus,RGS6 mediated voluntary running-induced rescue of impaired cognition and adult hippocampal neurogenesis in APP_(SWE) mice,identifying RGS6 in dentate gyrus neural precursor cells as a possible therapeutic target in Alzheimer’s disease.展开更多
Photonic platforms are gradually emerging as a promising option to encounter the ever-growing demand for artificial intelligence,among which photonic time-delay reservoir computing(TDRC)is widely anticipated.While suc...Photonic platforms are gradually emerging as a promising option to encounter the ever-growing demand for artificial intelligence,among which photonic time-delay reservoir computing(TDRC)is widely anticipated.While such a computing paradigm can only employ a single photonic device as the nonlinear node for data processing,the performance highly relies on the fading memory provided by the delay feedback loop(FL),which sets a restriction on the extensibility of physical implementation,especially for highly integrated chips.Here,we present a simplified photonic scheme for more flexible parameter configurations leveraging the designed quasi-convolution coding(QC),which completely gets rid of the dependence on FL.Unlike delay-based TDRC,encoded data in QC-based RC(QRC)enables temporal feature extraction,facilitating augmented memory capabilities.Thus,our proposed QRC is enabled to deal with time-related tasks or sequential data without the implementation of FL.Furthermore,we can implement this hardware with a low-power,easily integrable vertical-cavity surface-emitting laser for high-performance parallel processing.We illustrate the concept validation through simulation and experimental comparison of QRC and TDRC,wherein the simpler-structured QRC outperforms across various benchmark tasks.Our results may underscore an auspicious solution for the hardware implementation of deep neural networks.展开更多
文摘After billions of years of evolution,biological intelligence has converged on unrivalled energy efficiency and environmental adaptability.The human brain,for instance,is highly efficient in information transmission,consuming only about 20 W onaverage in a resting state[1,2].A key to this efficiency is that biological signal transduction and processing rely significantly on multi-ions as the signal carriers.Inspired by this paradigm.
基金Supported by the National Natural Science Foundation of China(Grant Nos.11961059,1210502)the University Innovation Project of Gansu Province(Grant No.2023B-062)the Gansu Province Basic Research Innovation Group Project(Grant No.23JRRA684).
文摘The goal of this paper is to investigate the long-time dynamics of solutions to a Kirchhoff type suspension bridge equation with nonlinear damping and memory term.For this problem we establish the well-posedness and existence of uniform attractor under some suitable assumptions on the nonlinear term g(u),the nonlinear damping f(u_(t))and the external force h(x,t).Specifically,the asymptotic compactness of the semigroup is verified by the energy reconstruction method.
基金supported by the National Natural Science Foundation of China(Grant No.62403486)。
文摘The highly dynamic nature,strong uncertainty,and coupled multiple safety constraints inherent in carrier aircraft recovery operations pose severe challenges for real-time decision-making.Addressing bolter scenarios,this study proposes an intelligent decision-making framework based on a deep long short-term memory Q-network.This framework transforms the real-time sequencing for bolter recovery problem into a partially observable Markov decision process.It employs a stacked long shortterm memory network to accurately capture the long-range temporal dependencies of bolter event chains and fuel consumption.Furthermore,it integrates a prioritized experience replay training mechanism to construct a safe and adaptive scheduling system capable of millisecond-level real-time decision-making.Experimental demonstrates that,within large-scale mass recovery scenarios,the framework achieves zero safety violations in static environments and maintains a fuel safety violation rate below 10%in dynamic scenarios,with single-step decision times at the millisecond level.The model exhibits strong generalization capability,effectively responding to unforeseen emergent situations—such as multiple bolters and fuel emergencies—without requiring retraining.This provides robust support for efficient carrier-based aircraft recovery operations.
基金supported by the National Natural Science Foundation of China(Nos.52473080,52403167 and 52173079)the Fundamental Research Funds for the Central Universities(Nos.xtr052023001 and xzy012023037)+1 种基金the Postdoctoral Research Project of Shaanxi Province(No.2024BSHSDZZ054)the Shaanxi Laboratory of Advanced Materials(No.2024ZY-JCYJ-04-12).
文摘Shape memory polymers used in 4D printing only had one permanent shape after molding,which limited their applications in requiring multiple reconstructions and multifunctional shapes.Furthermore,the inherent stability of the triazine ring structure within cyanate ester(CE)crosslinked networks after molding posed significant challenges for both recycling,repairing,and degradation of resin.To address these obstacles,dynamic thiocyanate ester(TCE)bonds and photocurable group were incorporated into CE,obtaining the recyclable and 3D printable CE covalent adaptable networks(CANs),denoted as PTCE1.5.This material exhibits a Young's modulus of 810 MPa and a tensile strength of 50.8 MPa.Notably,damaged printed PTCE1.5 objects can be readily repaired through reprinting and interface rejoining by thermal treatment.Leveraging the solid-state plasticity,PTCE1.5 also demonstrated attractive shape memory ability and permanent shape reconfigurability,enabling its reconfigurable 4D printing.The printed PTCE1.5 hinges and a main body were assembled into a deployable and retractable satellite model,validating its potential application as a controllable component in the aerospace field.Moreover,printed PTCE1.5 can be fully degraded into thiol-modified intermediate products.Overall,this material not only enriches the application range of CE resin,but also provides a reliable approach to addressing environmental issue.
基金supported by the National Key Research&Development Projects of China(Grant No.2022YFA1204100)National Natural Science Foundation of China(Grant No.62488201)+1 种基金CAS Project for Young Scientists in Basic Research(YSBR-003)the Innovation Program of Quantum Science and Technology(2021ZD0302700)。
文摘As a typical in-memory computing hardware design, nonvolatile ternary content-addressable memories(TCAMs) enable the logic operation and data storage for high throughout in parallel big data processing. However,TCAM cells based on conventional silicon-based devices suffer from structural complexity and large footprintlimitations. Here, we demonstrate an ultrafast nonvolatile TCAM cell based on the MoTe2/hBN/multilayergraphene (MLG) van der Waals heterostructure using a top-gated partial floating-gate field-effect transistor(PFGFET) architecture. Based on its ambipolar transport properties, the carrier type in the source/drain andcentral channel regions of the MoTe2 channel can be efficiently tuned by the control gate and top gate, respectively,enabling the reconfigurable operation of the device in either memory or FET mode. When working inthe memory mode, it achieves an ultrafast 60 ns programming/erase speed with a current on-off ratio of ∼105,excellent retention capability, and robust endurance. When serving as a reconfigurable transistor, unipolar p-typeand n-type FETs are obtained by adopting ultrafast 60 ns control-gate voltage pulses with different polarities.The monolithic integration of memory and logic within a single device enables the content-addressable memory(CAM) functionality. Finally, by integrating two PFGFETs in parallel, a TCAM cell with a high current ratioof ∼10^(5) between the match and mismatch states is achieved without requiring additional peripheral circuitry.These results provide a promising route for the design of high-performance TCAM devices for future in-memorycomputing applications.
基金supported by the National Key Research and Development Program of China under Grant No.2023YFB3308702the National Natural Science Foundation of China under Grant No.62302489the Strategy Priority Research Program of the Chinese Academy of Sciences under Grant No.XDA0360301.
文摘Supporting real-time and privacy-preserving learning at the edge is emerging as a critical trend,bringing forth substantial challenges for deep learning(DL)training in the context of limited GPU(graphic processing unit)memory.Recent work has sought to address the limitations by swapping tensors between GPU memory and CPU memory.Unfortunately,their tensor-based memory management encounters additional overhead since the swapped tensors do not align with the actual memory demands,resulting in decreased throughput.This paper introduces a vTensor-based memory management approach designed to mitigate memory swapping overhead.Virtualized tensors,dubbed vTensors,are used to finely align memory swapping amounts with real-time memory demands.Firstly,we introduce an abstraction layer that virtualizes coarse-grained tensors to multiple finer-grained vTensors.Secondly,we propose the Layered Graph Model(LGM)for analyzing vTensor mappings,which produces a memory swapping plan leveraged in the subsequent DL training iterations.Evaluations conducted on typical edge deep learning models illustrate that our approach surpasses prior work with a 15.60%increase in DL training throughput.
基金the National Natural Science Foundation of China(No.91632103)the Shanghai Education Commission Research and Innovation Program(No.2019-01-07-00-02-E00037)+2 种基金the Program of Shanghai Subject Chief Scientist(No.17XD1401700)the Higher Education Disciplinary Innovation Programthe“Eastern Scholar”Project。
文摘Working memory is a core cognitive function that supports goal-directed behavior and complex thought.We developed a spatial working memory and attention test on paired symbols(SWAPS)which has been proved to be a useful and valid tool for spatial working memory and attention studies in the fields of cognitive psychology,education,and psychiatry.The repeated administration of working memory capacity tests is common in clinical and research settings.Studies suggest that repeated cognitive tests may improve the performance scores also known as retest effects.The systematic investigation of retest effects in SWAPS is critical for interpreting scientific results,but it is still not fully developed.To address this,we recruited 77 college students aged 18–21 years and used SWAPS comprising 72 trials with different memory loads,learning time,and delay span.We repeated the test once a week for five weeks to investigate the retest effects of SWAPS.There were significant retest effects in the first two tests:the accuracy of the SWAPS tests significantly increased,and then stabilized.These findings provide useful information for researchers to appropriately use or interpret the repeated working memory tests.Further experiments are still needed to clarify the factors that mediate the retest effects,and find out the cognitive mechanism that influences the retest effects.
基金supported by grants from the Ministerio de Economia y Competitividad(BFU2013-43458-R)Junta de Andalucia(P12-CTS-1694 and Proyexcel-00422)to ZUK。
文摘Memory deficit,which is often associated with aging and many psychiatric,neurological,and neurodegenerative diseases,has been a challenging issue for treatment.Up till now,all potential drug candidates have failed to produce satisfa ctory effects.Therefore,in the search for a solution,we found that a treatment with the gene corresponding to the RGS14414protein in visual area V2,a brain area connected with brain circuits of the ventral stream and the medial temporal lobe,which is crucial for object recognition memory(ORM),can induce enhancement of ORM.In this study,we demonstrated that the same treatment with RGS14414in visual area V2,which is relatively unaffected in neurodegenerative diseases such as Alzheimer s disease,produced longlasting enhancement of ORM in young animals and prevent ORM deficits in rodent models of aging and Alzheimer’s disease.Furthermore,we found that the prevention of memory deficits was mediated through the upregulation of neuronal arbo rization and spine density,as well as an increase in brain-derived neurotrophic factor(BDNF).A knockdown of BDNF gene in RGS14414-treated aging rats and Alzheimer s disease model mice caused complete loss in the upregulation of neuronal structural plasticity and in the prevention of ORM deficits.These findings suggest that BDNF-mediated neuronal structural plasticity in area V2 is crucial in the prevention of memory deficits in RGS14414-treated rodent models of aging and Alzheimer’s disease.Therefore,our findings of RGS14414gene-mediated activation of neuronal circuits in visual area V2 have therapeutic relevance in the treatment of memory deficits.
基金supported by the National Natural Science Foundation of China(Nos.52201207 and 52271169)the Fundamental Research Funds for the Central University(No.3072024LJ1002).
文摘The active development of space industry necessitates the cre-ation of novel materials with unique properties,including shape memory alloys(SMAs).The development of ultra-high temperature SMAs(UHTSMAs)with operating temperatures above 400℃is a significant challenge[1-3].It is known that reversible thermoelas-tic martensitic transformation(MT)is the basis for shape mem-ory behavior[4].Currently,there are several systems in which MT temperatures meet the above requirements,for example,RuNb[5],HfPd[6],TiPd[7].
基金funded by Woosong University Academic Research 2024.
文摘This study investigates the application of Learnable Memory Vision Transformers(LMViT)for detecting metal surface flaws,comparing their performance with traditional CNNs,specifically ResNet18 and ResNet50,as well as other transformer-based models including Token to Token ViT,ViT withoutmemory,and Parallel ViT.Leveraging awidely-used steel surface defect dataset,the research applies data augmentation and t-distributed stochastic neighbor embedding(t-SNE)to enhance feature extraction and understanding.These techniques mitigated overfitting,stabilized training,and improved generalization capabilities.The LMViT model achieved a test accuracy of 97.22%,significantly outperforming ResNet18(88.89%)and ResNet50(88.90%),aswell as the Token to TokenViT(88.46%),ViT without memory(87.18),and Parallel ViT(91.03%).Furthermore,LMViT exhibited superior training and validation performance,attaining a validation accuracy of 98.2%compared to 91.0%for ResNet 18,96.0%for ResNet50,and 89.12%,87.51%,and 91.21%for Token to Token ViT,ViT without memory,and Parallel ViT,respectively.The findings highlight the LMViT’s ability to capture long-range dependencies in images,an areawhere CNNs struggle due to their reliance on local receptive fields and hierarchical feature extraction.The additional transformer-based models also demonstrate improved performance in capturing complex features over CNNs,with LMViT excelling particularly at detecting subtle and complex defects,which is critical for maintaining product quality and operational efficiency in industrial applications.For instance,the LMViT model successfully identified fine scratches and minor surface irregularities that CNNs often misclassify.This study not only demonstrates LMViT’s potential for real-world defect detection but also underscores the promise of other transformer-based architectures like Token to Token ViT,ViT without memory,and Parallel ViT in industrial scenarios where complex spatial relationships are key.Future research may focus on enhancing LMViT’s computational efficiency for deployment in real-time quality control systems.
文摘At the beginning of the 20^(th)century,German scientist Richard Semon introduced the term'engram'to describe the neural substrate implicated in the processes of memory formation and retrieval[1].The trace of the corresponding biophysical and biochemical changes in the brain responding to an external stimulus is called an engram,and understanding the physical manifestations of memory formation and recall remains a fundamental yet unresolved question[2].
基金supported by the National Natural Science Foundation of China(32371063,82341248,and 32071009)Guangdong Basic and Applied Basic Research Foundation(2024A1515011500)the Shenzhen Science and Technology Program(ZDSYS20200811142401005).
文摘Dear Editor,Sleep and memory are highly linked across species.Sleep gates and stabilizes memory,critical for memory processing.Insufficient sleep impairs cognition acutely/chronically,in vertebrates and invertebrates[1,2].While key elements are characterized[3,4],how a single molecule integrates sleep and memory remains unknown.
基金supported by the National Natural Science Foundation of China(Grant Nos.11633001,11920101003,and 12205222 for S.H.)the Key Program of the National Natural Science Foundation of China(Grant No.12433001)+1 种基金the Strate-gic Priority Research Program of the Chinese Academy of Sciences(Grant No.XDB23000000)the National Key Research and Development Program of China(Grant No.2021YFC2203001 for Z.C.Z.).
文摘The gravitational memory effect manifests gravitational nonlinearity,degenerate vacua,and asymptotic symmetries;its detection is considered challenging.We propose using a space-borne interferometer to detect memory signals from stellar-mass binary black holes(BBHs),typically targeted by ground-based detectors.We use DECIGO detector as an example.Over 5 years,DECIGO is estimated to detect approximately 2,036 memory signals(SNRs>3)from stellar-mass BBHs.Simulations used frequency-domain memory waveforms for direct SNR estimation.Predictions utilized a GWTC-3 constrained BBH population model(Power law+Peak mass,DEFAULT spin,Madau-Dickinson merger rate).The analysis used conservative lower merger rate limits and considered orbital eccentricity.The high detection rate stems from strong memory signals within DECIGO’s bandwidth and the abundance of stellar-mass BBHs.This substantial and conservative detection count enables statistical use of the memory effect for fundamental physics and astrophysics.DECIGO exemplifies that space interferometers may better detect memory signals from smaller mass binaries than their typical targets.Detectors in lower frequency bands are expected to find strong memory signals from∼10^(4)M⊙binaries.
基金supported by the National Natural Science Foundation of China(12202294)the Sichuan Science and Technology Program(2024NSFSC1346).
文摘Shape memory alloys(SMAs)and shape memory ceramics(SMCs)exhibit high recovery ability due to the martensitic transformation,which complicates the fracture mechanism of SMAs and SMCs.The phase field method,as a powerful numerical simulation tool,can efficiently resolve the microstructural evolution,multi-field coupling effects,and fracture behavior of SMAs and SMCs.This review begins by presenting the fundamental theoretical framework of the fracture phase field method as applied to SMAs and SMCs,covering key aspects such as the phase field modeling of martensitic transformation and brittle fracture.Subsequently,it systematically examines the phase field simulations of fracture behaviors in SMAs and SMCs,with particular emphasis on how crystallographic orientation,grain size,and grain boundary properties influence the crack propagation.Additionally,the interplay between martensite transformation and fracture mechanisms is analyzed to provide deeper insights into the material responses under mechanical loading.Finally,the review explores future prospects and emerging trends in phase field simulations of SMA and SMC fracture behavior,along with potential advancements in the fracture phase field method itself,including multi-physics coupling and enhanced computational efficiency for large-scale simulations.
基金supported in part by the National Natural Science Foundation of China (62422405, 62025111,62495100, 92464302)the STI 2030-Major Projects(2021ZD0201200)+1 种基金the Shanghai Municipal Science and Technology Major Projectthe Beijing Advanced Innovation Center for Integrated Circuits
文摘Computing-in-memory(CIM)has been a promising candidate for artificial-intelligent applications thanks to the absence of data transfer between computation and storage blocks.Resistive random access memory(RRAM)based CIM has the advantage of high computing density,non-volatility as well as high energy efficiency.However,previous CIM research has predominantly focused on realizing high energy efficiency and high area efficiency for inference,while little attention has been devoted to addressing the challenges of on-chip programming speed,power consumption,and accuracy.In this paper,a fabri-cated 28 nm 576K RRAM-based CIM macro featuring optimized on-chip programming schemes is proposed to address the issues mentioned above.Different strategies of mapping weights to RRAM arrays are compared,and a novel direct-current ADC design is designed for both programming and inference stages.Utilizing the optimized hybrid programming scheme,4.67×programming speed,0.15×power saving and 4.31×compact weight distribution are realized.Besides,this macro achieves a normalized area efficiency of 2.82 TOPS/mm2 and a normalized energy efficiency of 35.6 TOPS/W.
基金supported by the National Institutes of Health,Nos.AA025919,AA025919-03S1,and AA025919-05S1(all to RAF).
文摘Hippocampal neuronal loss causes cognitive dysfunction in Alzheimer’s disease.Adult hippocampal neurogenesis is reduced in patients with Alzheimer’s disease.Exercise stimulates adult hippocampal neurogenesis in rodents and improves memory and slows cognitive decline in patients with Alzheimer’s disease.However,the molecular pathways for exercise-induced adult hippocampal neurogenesis and improved cognition in Alzheimer’s disease are poorly understood.Recently,regulator of G protein signaling 6(RGS6)was identified as the mediator of voluntary running-induced adult hippocampal neurogenesis in mice.Here,we generated novel RGS6fl/fl;APP_(SWE) mice and used retroviral approaches to examine the impact of RGS6 deletion from dentate gyrus neuronal progenitor cells on voluntary running-induced adult hippocampal neurogenesis and cognition in an amyloid-based Alzheimer’s disease mouse model.We found that voluntary running in APP_(SWE) mice restored their hippocampal cognitive impairments to that of control mice.This cognitive rescue was abolished by RGS6 deletion in dentate gyrus neuronal progenitor cells,which also abolished running-mediated increases in adult hippocampal neurogenesis.Adult hippocampal neurogenesis was reduced in sedentary APP_(SWE) mice versus control mice,with basal adult hippocampal neurogenesis reduced by RGS6 deletion in dentate gyrus neural precursor cells.RGS6 was expressed in neurons within the dentate gyrus of patients with Alzheimer’s disease with significant loss of these RGS6-expressing neurons.Thus,RGS6 mediated voluntary running-induced rescue of impaired cognition and adult hippocampal neurogenesis in APP_(SWE) mice,identifying RGS6 in dentate gyrus neural precursor cells as a possible therapeutic target in Alzheimer’s disease.
基金National Natural Science Foundation of China(62171305,62405206,62004135,62001317,62111530301)Natural Science Foundation of Jiangsu Province(BK20240778,BK20241917)+3 种基金State Key Laboratory of Advanced Optical Communication Systems and Networks,China(2023GZKF08)China Postdoctoral Science Foundation(2024M752314)Postdoctoral Fellowship Program of CPSF(GZC20231883)Innovative and Entrepreneurial Talent Program of Jiangsu Province(JSSCRC2021527).
文摘Photonic platforms are gradually emerging as a promising option to encounter the ever-growing demand for artificial intelligence,among which photonic time-delay reservoir computing(TDRC)is widely anticipated.While such a computing paradigm can only employ a single photonic device as the nonlinear node for data processing,the performance highly relies on the fading memory provided by the delay feedback loop(FL),which sets a restriction on the extensibility of physical implementation,especially for highly integrated chips.Here,we present a simplified photonic scheme for more flexible parameter configurations leveraging the designed quasi-convolution coding(QC),which completely gets rid of the dependence on FL.Unlike delay-based TDRC,encoded data in QC-based RC(QRC)enables temporal feature extraction,facilitating augmented memory capabilities.Thus,our proposed QRC is enabled to deal with time-related tasks or sequential data without the implementation of FL.Furthermore,we can implement this hardware with a low-power,easily integrable vertical-cavity surface-emitting laser for high-performance parallel processing.We illustrate the concept validation through simulation and experimental comparison of QRC and TDRC,wherein the simpler-structured QRC outperforms across various benchmark tasks.Our results may underscore an auspicious solution for the hardware implementation of deep neural networks.
