检索结果-南通市图书馆

Ensemble successor representations for task generalization in offline-to-online reinforcement learning

Science China(Information Sciences) 2024年第7期67卷 240-255页

作者： Changhong WANG Xudong YU chenjia bai Qiaosheng ZHANG Zhen WANG Space Control and Inertial Technology Research Center Harbin Institute of Technology Shanghai Artificial Intelligence Laboratory Shenzhen Research Institute of Northwestern Polytechnical University School of Cybersecurity Northwestern Polytechnical University

In reinforcement learning(RL), training a policy from scratch with online experiences can be inefficient because of the difficulties in exploration. Recently, offline RL provides a promising solution by giving an initialized offline policy, which can be refined through online interactions. However, existing approaches primarily perform offline and online learning in the same task, without considering the task generalization problem in offline-to-online adaptation. In real-world applications, it is common that we only have an offline dataset from a specific task while aiming for fast online-adaptation for several tasks. To address this problem, our work builds upon the investigation of successor representations for task generalization in online RL and extends the framework to incorporate offline-to-online learning. We demonstrate that the conventional paradigm using successor features cannot effectively utilize offline data and improve the performance for the new task by online fine-tuning. To mitigate this, we introduce a novel methodology that leverages offline data to acquire an ensemble of successor representations and subsequently constructs ensemble Q functions. This approach enables robust representation learning from datasets with different coverage and facilitates fast adaption of Q functions towards new tasks during the online fine-tuning *** empirical evaluations provide compelling evidence showcasing the superior performance of our method in generalizing to diverse or even unseen tasks.

关键词： offline reinforcement learning online fine-tuning task generalization successor representations ensembles

来源：

同方期刊数据库评论

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

基于TD-error自适应校正的深度Q学习主动采样方法

引用

计算机研究与发展 2019年第2期56卷 262-280页

作者：白辰甲刘鹏赵巍唐降龙哈尔滨工业大学计算机科学与技术学院模式识别与智能系统研究中心哈尔滨150001

强化学习中智能体与环境交互的成本较高.针对深度Q学习中经验池样本利用效率的问题,提出基于TD-error自适应校正的主动采样方法.深度Q学习训练中样本存储优先级的更新滞后于Q网络参数的更新,存储优先级不能准确反映经验池中样本TD-erro... 详细信息

强化学习中智能体与环境交互的成本较高.针对深度Q学习中经验池样本利用效率的问题,提出基于TD-error自适应校正的主动采样方法.深度Q学习训练中样本存储优先级的更新滞后于Q网络参数的更新,存储优先级不能准确反映经验池中样本TD-error的真实分布.提出的TD-error自适应校正主动采样方法利用样本回放周期和Q网络状态建立优先级偏差模型,估计经验池中样本的真实优先级.在Q网络迭代中使用校正后的优先级选择样本,偏差模型在学习过程中分段更新.分析了Q网络学习性能与偏差模型阶数和模型更新周期之间的依赖关系,并对算法复杂度进行了分析.方法在Atari 2600平台进行了实验,结果表明,使用TD-error自适应校正的主动采样方法选择样本提高了智能体的学习速度,减少了智能体与环境的交互次数,同时改善了智能体的学习效果,提升了最优策略的质量.

关键词：样本优先级 TD-error校正自适应主动采样深度Q学习强化学习

来源：

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

大模型驱动的具身智能:发展与挑战

引用

中国科学：信息科学 2024年第9期54卷 2035-2082页

作者：白辰甲许华哲李学龙中国电信人工智能研究院(TeleAI) 上海200232 中国电信人工智能研究院(TeleAI) 北京100033 清华大学交叉信息学院北京100084

大模型驱动的具身智能是涵盖人工智能、机器人学和认知科学的交叉领域,重点研究如何将大模型的感知、推理和逻辑思维能力与具身智能相结合,提升现有模仿学习、强化学习、模型预测控制等具身智能框架的数据效率和泛化能力.近年来,随着大... 详细信息

大模型驱动的具身智能是涵盖人工智能、机器人学和认知科学的交叉领域,重点研究如何将大模型的感知、推理和逻辑思维能力与具身智能相结合,提升现有模仿学习、强化学习、模型预测控制等具身智能框架的数据效率和泛化能力.近年来,随着大模型能力的不断提升,以及具身智能中示教数据、仿真平台、任务集合的不断完善,大模型和具身智能的结合将成为人工智能的下一个浪潮,有望成为人工智能迈向实体机器人的重要突破口.本文围绕大模型驱动的具身智能这一研究领域,从3个方面进行了系统的调研、分析和展望.首先,回顾了大模型和具身智能的相关技术背景,以及具身智能现有的学习框架.其次,按照大模型赋能具身智能的方式,将现有研究分为大模型驱动的环境感知、大模型驱动的任务规划、大模型驱动的基础策略、大模型驱动的奖励函数、大模型驱动的数据生成等5类范式.最后,总结了大模型驱动的具身智能中存在的挑战,对可行的技术路线进行展望,为相关研究人员提供参考,进一步推动国家人工智能发展战略.

关键词：具身智能大模型环境感知任务规划基础策略

来源：

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

Boosting bidirectional sulfur conversion enabled by introducing boron-doped atoms and phosphorus vacancies in Ni2P for lithium-sulfur batteries

引用

Journal of Energy Chemistry 2025年第1期100卷 760-769页

作者： Lin Peng Yu bai Hang Li Meixiu Qu Zhenhua Wang Kening Sun Advanced Research Institute of Multidisciplinary Science Beijing Institute of Technology Beijing Key Laboratory of Chemical Power Source and Green Catalysis School of Chemistry and Chemical Engineering Beijing Institute of Technology

Lithium-sulfur （Li-S） batteries have gained great attention due to the high theoretical energy density and low cost,yet their further commercialization has been obstructed by the notorious shuttle effect and sluggish redox ***,we supply a strategy to optimize the electron structure of Ni2P by concurrently introducing B-doped atoms and P vacancies in Ni2P （Vp-B-Ni2P）,thereby enhancing the bidirectional sulfur *** study indicates that the simultaneous introduction of B-doped atoms and P vacancies in Ni2P causes the redistribution of electron around Ni atoms,bringing about the upward shift of d-band center of Ni atoms and effective d-p orbital hybridization between Ni atoms and sulfur species,thus strengthening the chemical anchoring for lithium polysulfides （LiPSs） as well as expediting the bidirectional conversion kinetics of sulfur ***,theoretical calculations reveal that the incorporation of B-doped atoms and P vacancies in Ni2P selectively promotes Li2S dissolution and nucleation ***,the Li-S batteries with Vp-B-Ni2P-separators present outstanding rate ability of 777 m A h g-1at 5 C and high areal capacity of 8.03 mA h cm-2under E/S of 5μL mg-1and sulfur loading of 7.20 mg *** work elucidates that introducing heteroatom and vacancy in metal phosphide collaboratively regulates the electron structure to accelerate bidirectional sulfur conversion.

关键词： Lithium sulfur batteries

来源：

同方期刊数据库评论

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

NIR photothermal enhancement to achieve high-efficiency nitrogen reduction to ammonia by polyoxometalates@Fe-polydopamine

引用

Nano Research 2025年第1期18卷 118-128页

作者： Xiaofei Chen Lixin Wu Feng bai Key Laboratory for Special Functional Materials of Ministry of Education National&Local Joint Engineering Research Center for High-Efficiency Display and Lighting TechnologySchool of Nanoscience and Materials EngineeringHenan UniversityKaifeng 475004China State Key Laboratory of Supramolecular Structure and Materials College of ChemistryJilin UniversityChangchun 130012China

The development of atmospheric pressure N_(2)reduction to NH_(3)is attracting much attention in green chemistry,yet it is still a challenge to obtain satisfactory activity under mild ***,an efficient near-infrared(NIR)photothermal catalysis reduction of N_(2)constitutes an occurrence is *** or without V-substitute polyoxometalates(POMs)loaded on the surface of Fechelated polydopamine(Fe-PDA)photothermal support through the electrostatic interactions,NIR photothermal catalysts POMs@Fe-PDA are *** induction of“FeV”cofactor facilitates electron transfer between V(V)/V(IV)&Fe(III)/Fe(II)and N_(2),thereby activating N_(2)*** synergy between the catalytic activity of V-POMs and the local NIR photothermal effect of Fe-PDA dramatically enhances N_(2)***,PMo_(10)V_(2)@Fe-PDA exhibits a significantly enhanced NH_(3)production rate of 181.1μmol·LN^(-1)with a turnover frequency of 1006.1 mmol·MN^(-1)·hN^(-1)under 808 nm NIR laser radiation,being the highest values reported at atmospheric *** expect that this work could provide an alternative approach for photothermal catalysis N_(2)reduction under mild conditions.

关键词： polyoxometalate near-infrared photothermal catalysis N_(2)reduction mild condition

来源：

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Synergistic effect of bonding heterogeneity and phonon localization in introducing excellent thermoelectric properties in layered heteroanionic NdZnSbO material

引用

Journal of Energy Chemistry 2025年第1期100卷 259-272页

作者： Da Wan Shuwei Tang Shulin bai Xiaodong Li Peng Ai Wanrong Guo Tuo Zheng College of Materials Science and Engineering Liaoning Technical University Faculty of Chemistry Northeast Normal University

Layered rare-earth metal oxides, harnessing the dual properties of oxides and two-dimensional layered materials, exhibit remarkable thermal stability and quantum confinement effects. Therefore, this work adopts the first-principles calculation combined with the Boltzmann transport theory to predict the thermoelectric properties of NdZnSbO compound. The coexistence of weak interlayer van der Waals interactions, robust intralayer ionic bonding, and partial covalent bonding leads to remarkable bonding heterogeneity, which engenders pronounced phonon scattering and imposes constraints on thermal transport along the out-of-plane direction. The weakened chemical bonds induced by the antibonding states, together with the rattling-like behavior of the Zn atom, culminate in the profound anharmonicity in the layered NdZnSbO compound. The weakening bond and heavy element contribute to the softness of phonon modes, which significantly diminishes the phonon group velocity. The redistribution-dominated four-phonon scattering process spans a large optical gap, which effectively reduces the lattice thermal conductivity. The NdZnSbO compound exhibits direct semiconductor characteristic with a bandgap of 0.73 e V by adopting the Heyd-Scuseria-Ernzerhof(HSE06) functional in combination with spin–orbit coupling(SOC) effect. The multi-valley feature of NdZnSbO compound augur favorably for band degeneracy, thus amplifying the power factor. Consequently, an optimal figure-of-merit(ZT) of 3.40 at 900 K is achieved for the n-type NdZnSbO compound. The present study delves deeply insights into the origins for the low thermal conductivity of NdZnSbO compound and proposes an optimization scheme to enhance overall thermoelectric performance.

关键词： Bond strength (chemical)

来源：

同方期刊数据库评论

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

New exploration on pathogenesis and early diagnosis of gestational diabetes

引用

World Journal of Clinical Cases 2025年第1期13卷 1-5页

作者： Hua bai Department of Neurology The Third Affiliated Hospital of Guizhou Medical UniversityDuyun 558099Guizhou ProvinceChina

Gestational diabetes mellitus(GDM)refers to varying degrees of abnormal glucose metabolism that occur during pregnancy and excludes patients pre-viously diagnosed with *** is a unique among the four subtypes of diabetes classified by the international World Health Organization *** GDM patients constitute a small proportion of the total number of diabetes cases,the incidence of GDM has risen significantly over the past decade,posing substantial risk to pregnant women and ***,it warrants considerable *** pathogenesis of GDM is generally considered to resemble that of type II diabetes,though it may have distinct *** blood biochemical proteins in the context of GDM can help elucidate its pathogenesis,thereby facilitating more effective prevention and management *** article reviews this critical clinical issue to enhance the medical community's sufficient understanding of GDM.

关键词： Gestational diabetes mellitus Pathogenesis Proteomics RBP4 ANGPTL8

来源：

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Searching for new drugs to treat Alzheimer’s disease dementia through multiple pathways

引用

World Journal of Clinical Cases 2025年第1期13卷 22-28页

作者： Hua bai Xiao-Feng Feng Department of Neurology The Third Affiliated Hospital of Guizhou Medical UniversityDuyun 558099Guizhou ProvinceChina Department of Neurology Wulong Branch of the People's Hospital Affiliated to Chongqing UniversityWulong 408500ChongqingChina Department of Neurology Guizhou Medical UniversityGuiyang 550004Guizhou ProvinceChina

Dementia is a group of diseases,including Alzheimer's disease(AD),vascular dementia,Lewy body dementia,frontotemporal dementia,Parkinson's disease dementia,metabolic dementia and toxic *** treatment of dementia mainly includes symptomatic treatment by controlling the primary disease and accompanying symptoms,nutritional support therapy for repairing nerve cells,psychological auxiliary treatment,and treatment that improves cognitive function through *** them,drug therapy to improve cognitive function is *** review focuses on introducing and commenting on some recent progress in exploring drugs to improve cognitive function,especially the new progress in drug treatment for *** mainly discuss the opportunities and challenges in finding and developing new therapeutic drugs from the aspects of acetylcholinesterase,N-methyl-D-aspartate glutamate receptor,amyloid protein,tau protein and chronic immune inflammation.

关键词： Alzheimer's disease Dementia Cholinesterase inhibitor Beta amyloid Tau protein

来源：

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Brain region-specific roles of brain-derived neurotrophic factor in social stress-induced depressive-like behavior

引用

Neural Regeneration Research 2025年第1期20卷 159-173页

作者： Man Han Deyang Zeng Wei Tan Xingxing Chen Shuyuan bai Qiong Wu Yushan Chen Zhen Wei Yufei Mei Yan Zeng Brain Science and Advanced Technology Institute Wuhan University of Science and TechnologyWuhanHubei ProvinceChina Geriatric Hospital Affiliated to Wuhan University of Science and Technology WuhanHubei ProvinceChina School of Public Health Wuhan University of Science and TechnologyWuhanHubei ProvinceChina

Brain-derived neurotrophic factor is a key factor in stress adaptation and avoidance of a social stress behavioral *** studies have shown that brain-derived neurotrophic factor expression in stressed mice is brain region–specific,particularly involving the corticolimbic system,including the ventral tegmental area,nucleus accumbens,prefrontal cortex,amygdala,and *** how brain-derived neurotrophic factor participates in stress processing in different brain regions will deepen our understanding of social stress *** this review,we discuss the expression and regulation of brain-derived neurotrophic factor in stress-sensitive brain regions closely related to the pathophysiology of *** focused on associated molecular pathways and neural circuits,with special attention to the brain-derived neurotrophic factor–tropomyosin receptor kinase B signaling pathway and the ventral tegmental area–nucleus accumbens dopamine *** determined that stress-induced alterations in brain-derived neurotrophic factor levels are likely related to the nature,severity,and duration of stress,especially in the above-mentioned brain regions of the corticolimbic ***,BDNF might be a biological indicator regulating stress-related processes in various brain regions.

关键词： amygdala chronic mild stress chronic social defeat stress corticolimbic system depression hippocampus medial prefrontal cortex nucleus accumbens social stress models ventral tegmental area

来源：

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

吐鲁番9个产地无核白葡萄干品质及多酚代谢组分析

引用

食品科学 2025年第2期46卷 156-163页

作者：户金鸽白世践新疆维吾尔自治区葡萄瓜果研究所新疆鄯善838200

以吐鲁番9个产地无核白葡萄干为研究对象,分析比较不同产区间葡萄干品质、抗氧化活性差异,采用超高效液相色谱-串联质谱法测定葡萄干酚类组分和含量。结果表明,平均葡萄干单粒质量和果形指数最大的是连木沁,分别为0.81 g、2.74,平均单... 详细信息

以吐鲁番9个产地无核白葡萄干为研究对象,分析比较不同产区间葡萄干品质、抗氧化活性差异,采用超高效液相色谱-串联质谱法测定葡萄干酚类组分和含量。结果表明,平均葡萄干单粒质量和果形指数最大的是连木沁,分别为0.81 g、2.74,平均单粒质量和果形指数最小的是高昌区,分别为0.46 g、2.27。园艺场葡萄干固酸比最高,为29.34,七克台葡萄干固酸比最低,为25.74。连木沁葡萄干a*值最低,ΔE较低,分别为2.56和44.94,表明其葡萄干颜色最绿;园艺场葡萄干a*值较低,ΔE最低,分别为2.94和44.19,这两个产区内葡萄干颜色偏绿且颜色一致性最好。对6个产区无核白葡萄干检测到的518种单体酚进行分类,可分为5大类:黄酮275种,占53.09%;酚酸136种,占26.25%;木脂素和香豆素60种,占11.58%;单宁19种,占3.67%;其他类(茋类)28种,占5.41%。鲁克沁和高昌区、七克台和胜金乡、园艺场和连木沁3组共筛选到246个差异代谢物,分别包含124、144、66个差异代谢物,各有54、92、22个特有差异代谢物,这3组之间有10个共同差异代谢物。上述结果表明,连木沁和园艺场可作为无核白葡萄干的优势产区。

关键词：吐鲁番葡萄干品质酚类代谢组

来源：

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

欢迎您,

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

在线全文

在线全文

在线全文

在线全文

在线全文

在线全文

在线全文

在线全文

在线全文

在线全文

请选择保存的检索档案：

请选择收藏分类：

通借通还

欢迎您,

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

在线全文

在线全文

在线全文

在线全文

在线全文

在线全文

在线全文

在线全文

在线全文

在线全文

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：