CAREER: Towards Real-world Reinforcement Learning
职业:走向现实世界的强化学习
基本信息
- 批准号:2339395
- 负责人:
- 金额:$ 60万
- 依托单位:
- 依托单位国家:美国
- 项目类别:Continuing Grant
- 财政年份:2024
- 资助国家:美国
- 起止时间:2024-03-01 至 2029-02-28
- 项目状态:未结题
- 来源:
- 关键词:
项目摘要
Reinforcement learning (RL) is one of the most important paradigms for modeling data-driven decision-making. Recent years have witnessed several empirical successes of RL, such as RL agents that outperform humans in video and board games. However, many empirical RL algorithms today often require many training examples to learn and can produce unreliable solutions (solutions that exhibit catastrophic failures, for example). While these issues are typically not problematic when training RL agents in simulators, they pose significant difficulties when deploying RL to real-world problems where data (including human feedback) is expensive, and reliability is essential. The main novelty of this project will be the development of new RL algorithms that can learn efficiently (from as few training data points as possible) and reliably (avoid catastrophic failures with high probability). The development of such RL algorithms can expand the applications of RL systems from simulation to real-world applications where data is expensive to collect and safety is critical. In autonomous driving, the developed technologies can make self-driving cars adapt to new road conditions safely by making fewer mistakes. In generative Artificial Intelligence (AI), efficient and reliable RL algorithms that can learn from rich human feedback will enable better human-AI alignment, making AI systems improve reliably and safely under human guidance.The main research goal of this project is to enable real-world RL by advancing RL techniques, theoretically and empirically. The critical innovation in the project is to develop safe and efficient RL algorithms by leveraging specific problem structures and rich human feedback. The project has three main thrusts. First, the project will establish risk-averse RL algorithms that are provably correct and scalable to high dimensional data. Second, the project will develop RL algorithms that can leverage common problem-specific structures for improved sample efficiency. Third, the project will create new algorithms for RL with rich feedback beyond scalar rewards (including preference-based feedback and positive demonstrations). In addition to the proposed work on algorithmic advancements, the project will focus on their deployment to real-world problems, including database query optimization and optimizing generative models such as Large Language Models and Diffusion Models.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.
强化学习(RL)是数据驱动决策建模最重要的范例之一。近年来,强化学习在实践中取得了一些成功,例如强化学习智能体在视频和棋盘游戏中的表现优于人类。然而,当今许多经验强化学习算法通常需要许多训练样本来学习,并且可能会产生不可靠的解决方案(例如,出现灾难性故障的解决方案)。虽然在模拟器中训练 RL 代理时,这些问题通常不会成为问题,但在将 RL 部署到数据(包括人类反馈)昂贵且可靠性至关重要的现实问题中时,它们会带来很大的困难。该项目的主要新颖之处在于开发新的强化学习算法,该算法可以高效(从尽可能少的训练数据点)和可靠地学习(避免高概率的灾难性故障)。这种 RL 算法的开发可以将 RL 系统的应用从模拟扩展到现实世界的应用,在现实世界中,数据收集成本昂贵且安全性至关重要。在自动驾驶方面,发达的技术可以使自动驾驶汽车安全地适应新的路况,减少错误。在生成人工智能(AI)中,高效可靠的强化学习算法可以从丰富的人类反馈中学习,从而实现更好的人类与人工智能的协调,使人工智能系统在人类指导下可靠、安全地改进。该项目的主要研究目标是通过在理论和经验上推进强化学习技术来实现现实世界的强化学习。该项目的关键创新是通过利用特定的问题结构和丰富的人类反馈来开发安全高效的强化学习算法。该项目有三个主要目标。首先,该项目将建立规避风险的强化学习算法,该算法可被证明是正确的并且可扩展到高维数据。其次,该项目将开发强化学习算法,该算法可以利用常见的特定问题结构来提高样本效率。第三,该项目将为强化学习创建新的算法,提供除标量奖励之外的丰富反馈(包括基于偏好的反馈和积极的示范)。除了拟议的算法改进工作外,该项目还将重点关注它们对现实世界问题的部署,包括数据库查询优化和优化生成模型,例如大型语言模型和扩散模型。该奖项反映了 NSF 的法定使命,并通过使用基金会的智力价值和更广泛的影响审查标准进行评估,被认为值得支持。
项目成果
期刊论文数量(0)
专著数量(0)
科研奖励数量(0)
会议论文数量(0)
专利数量(0)
数据更新时间:{{ journalArticles.updateTime }}
{{
item.title }}
{{ item.translation_title }}
- DOI:
{{ item.doi }} - 发表时间:
{{ item.publish_year }} - 期刊:
- 影响因子:{{ item.factor }}
- 作者:
{{ item.authors }} - 通讯作者:
{{ item.author }}
数据更新时间:{{ journalArticles.updateTime }}
{{ item.title }}
- 作者:
{{ item.author }}
数据更新时间:{{ monograph.updateTime }}
{{ item.title }}
- 作者:
{{ item.author }}
数据更新时间:{{ sciAawards.updateTime }}
{{ item.title }}
- 作者:
{{ item.author }}
数据更新时间:{{ conferencePapers.updateTime }}
{{ item.title }}
- 作者:
{{ item.author }}
数据更新时间:{{ patent.updateTime }}
Wen Sun其他文献
No-Regret Safe Learning for Online Nonlinear Control with Control Barrier Functions
具有控制屏障函数的在线非线性控制的无悔安全学习
- DOI:
- 发表时间:
2021 - 期刊:
- 影响因子:0
- 作者:
Wenhao Luo;Wen Sun;Ashish Kapoor - 通讯作者:
Ashish Kapoor
Dipole-induced modulation of effective work function of metal gate in junctionless FETs
无结 FET 中金属栅极有效功函数的偶极子感应调制
- DOI:
10.1063/1.5143771 - 发表时间:
2020 - 期刊:
- 影响因子:1.6
- 作者:
Xinhe Wang;Zhigang Zhang;Jianshi Tang;B. Gao;Wen Sun;Feng Xu;Huaqiang Wu;He Qian - 通讯作者:
He Qian
Large and reversible elastocaloric effect in dual-phase Ni54Fe19Ga27 superelastic alloys
双相 Ni54Fe19Ga27 超弹性合金中大且可逆的弹热效应
- DOI:
10.1063/1.4921531 - 发表时间:
2015-05 - 期刊:
- 影响因子:4
- 作者:
Yang Xu;Binfeng Lu;Wen Sun;Aru Yan;Jian Liu - 通讯作者:
Jian Liu
In Utero Exposure to Fine Particles Decreases Early Birth Weight of Rat Offspring and TLR4/NF-κB Expression in Lungs
子宫内暴露于细颗粒会降低大鼠后代的早期出生体重和肺部 TLR4/NF-κB 表达
- DOI:
10.1021/acs.chemrestox.0c00056 - 发表时间:
2021 - 期刊:
- 影响因子:0
- 作者:
Wenting Tang;Zhongjun Li;Yaoguang Huang;Lili Du;Chuangyu Wen;Wen Sun;Zhiqiang Yu;Suran Huang;Dunjin Chen - 通讯作者:
Dunjin Chen
Synchronization criterions between two identical or different fractional order chaotic systems
两个相同或不同分数阶混沌系统之间的同步准则
- DOI:
- 发表时间:
2011 - 期刊:
- 影响因子:0
- 作者:
Yuhua Xu;Wuneng Zhou;Jian'an Fang;Lin Pan;Wen Sun - 通讯作者:
Wen Sun
Wen Sun的其他文献
{{
item.title }}
{{ item.translation_title }}
- DOI:
{{ item.doi }} - 发表时间:
{{ item.publish_year }} - 期刊:
- 影响因子:{{ item.factor }}
- 作者:
{{ item.authors }} - 通讯作者:
{{ item.author }}
{{ truncateString('Wen Sun', 18)}}的其他基金
RI: Small: Towards Provably Efficient Representation Learning in Reinforcement Learning via Rich Function Approximation
RI:小:通过丰富函数逼近实现强化学习中可证明有效的表示学习
- 批准号:
2154711 - 财政年份:2022
- 资助金额:
$ 60万 - 项目类别:
Standard Grant
相似海外基金
CAREER: Towards Safety-Critical Real-Time Systems with Learning Components
职业:迈向具有学习组件的安全关键实时系统
- 批准号:
2340171 - 财政年份:2024
- 资助金额:
$ 60万 - 项目类别:
Continuing Grant
Towards a real implementation of quantum network systems
迈向量子网络系统的真正实现
- 批准号:
24K07485 - 财政年份:2024
- 资助金额:
$ 60万 - 项目类别:
Grant-in-Aid for Scientific Research (C)
CAREER: Towards Fairness in the Real World under Generalization, Privacy and Robustness Challenges
职业:在泛化、隐私和稳健性挑战下实现现实世界的公平
- 批准号:
2339198 - 财政年份:2024
- 资助金额:
$ 60万 - 项目类别:
Continuing Grant
CRII: SCH: Towards Smart Patient Flow Management: Real-time Inpatient Length of Stay Modeling and Prediction
CRII:SCH:迈向智能患者流程管理:实时住院患者住院时间建模和预测
- 批准号:
2246158 - 财政年份:2023
- 资助金额:
$ 60万 - 项目类别:
Standard Grant
Towards Real-Time Fine-Grained Tracking in Distributed Large-Scale RF Tag Systems
实现分布式大规模射频标签系统中的实时细粒度跟踪
- 批准号:
2225337 - 财政年份:2023
- 资助金额:
$ 60万 - 项目类别:
Standard Grant
Collaborative Research: Towards Engaged, Personalized and Transferable Learning of Secure Programming by Leveraging Real-World Security Vulnerabilities
协作研究:利用现实世界的安全漏洞实现安全编程的参与式、个性化和可转移学习
- 批准号:
2235976 - 财政年份:2023
- 资助金额:
$ 60万 - 项目类别:
Standard Grant
Towards Real-world Continual Learning on Unrestricted Task Steams
在不受限制的任务流上实现现实世界的持续学习
- 批准号:
DE230101591 - 财政年份:2023
- 资助金额:
$ 60万 - 项目类别:
Discovery Early Career Researcher Award
Collaborative Research: Towards Engaged, Personalized and Transferable Learning of Secure Programming by Leveraging Real-World Security Vulnerabilities
协作研究:利用现实世界的安全漏洞实现安全编程的参与式、个性化和可转移学习
- 批准号:
2235224 - 财政年份:2023
- 资助金额:
$ 60万 - 项目类别:
Standard Grant
Towards real-time precision crop management: automatic during-harvest assessment of apple yield
实现实时精准作物管理:苹果产量收获期间自动评估
- 批准号:
10080669 - 财政年份:2023
- 资助金额:
$ 60万 - 项目类别:
Grant for R&D
Towards the real-time analysis of phase transition
迈向相变的实时分析
- 批准号:
23K17687 - 财政年份:2023
- 资助金额:
$ 60万 - 项目类别:
Grant-in-Aid for Challenging Research (Exploratory)