AI Institute for Advances in Optimization

这家NSF人工智能(AI)优化进展研究所旨在通过融合AI和数学优化(MO)来实现这两个领域都无法独立实现的突破，从而实现大规模自动化决策的范式转变。该研究所是由能源、物流和供应链、复原力和可持续性以及电路设计和控制方面的社会挑战推动的。特别是，该研究所将帮助提供下一代控制和优化算法，用于运行具有分布式发电和需求响应的电网，以及规划和调度大规模、绿色和有弹性的供应链。该研究所还将开创新的AI&Amp；MO方法，用于设计新的混合信号集成电路，大幅缩短开发时间，并用于运营可持续的城市环境。为了解决日益扩大的就业机会差距，该研究所将提供一个创新的纵向教育和劳动力发展计划，最初的重点是佐治亚州历史上的黑人高中和大学，以及加利福尼亚州为拉美裔服务的高中和大学。该项目沿着“工程学中的人工智能之路”进行组织，并寻求与高中和社区大学建立长期合作伙伴关系，以改变针对服务不足学生的人工智能教育和研究。这些途径将给现有的课程带来一步的改变，并在工程学科和社会挑战的背景下教授人工智能和人工智能。该研究所将与国家实验室和行业合作伙伴开发实习计划，并建立一个强大、欢迎和包容的社区，在其中将突出社会流动机会和人工智能技术的社会影响。该研究所汇集了人工智能和优化领域的多学科团队和最终使用案例中的领域专家，汇集了佐治亚理工学院、加州大学伯克利分校、南加州大学克拉克·亚特兰大大学、斯佩尔曼学院和德克萨斯大学阿林顿分校。为了大规模转变决策，该研究所从优化解决方案转向智能代理，这些代理可以预测和量化不确定性、推理和优化，不断学习，并进行协调和协作。它统一了作为人工智能和运筹学(OR)核心的数据驱动和模型驱动的方法。它的方法论推动力包括学习优化的新一代混合优化求解器，将预测和决策紧密结合的端到端学习和优化，以及基于组合优化的新型机器学习方法。为了在大规模范围内学习和优化，该研究所将在紧凑表示、数据压缩和概率建模方面做出创新。为了在工程学科中经常出现的不确定和多智能体环境中实现安全和可扩展的决策，该研究所将在强化学习、分散优化和大规模数据驱动优化方面设计新方法。重要的是，为了确保这些科学进步服务于社会利益，横向推力将在复杂系统设计中从一开始就将伦理和价值观整合到设计和运营中。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

This NSF Artificial Intelligence (AI) Research Institute for Advances in Optimization aims at delivering a paradigm shift in automated decision-making at massive scales by fusing AI and Mathematical Optimization (MO) to achieve breakthroughs that neither field can achieve independently. The Institute is driven by societal challenges in energy, logistics and supply chains, resilience and sustainability, and circuit design and control. In particular, the Institute will help deliver the next generation of control and optimization algorithms for operating electric grids with distributed generation and demand response, and for planning and scheduling large-scale, green, and resilient supply chains. The Institute will also pioneer novel AI&MO methods for designing new mixed-signal integrated circuits, dramatically reducing development time, and for operating sustainable urban environments. To address the widening gap in job opportunities, the Institute will deliver an innovative longitudinal education and workforce development program with an initial focus on historically black high schools and colleges in Georgia, as well as Hispanic-serving high-schools and colleges in California. The program is organized along “pathways for AI in engineering” and seeks long-term partnerships with high schools and community colleges to transform AI education and research for underserved students. These pathways will bring a step change to existing programs and teach AI&MO in the context of engineering disciplines and societal challenges. The Institute will develop internship programs with national laboratories and industrial partners, and build a strong, welcoming, and inclusive community where social mobility opportunities and the societal impact of AI technologies will be highlighted. The Institute assembles a multi-disciplinary team in artificial intelligence and optimization and domain experts in the end-use cases, bringing together the Georgia Institute of Technology, the University of California at Berkeley, the University of Southern California, Clark Atlanta University, Spelman College, and the University of Texas at Arlington.To transform decision-making at massive scales, the Institute moves from optimization solutions to intelligent agents that predict and quantify uncertainty, reason and optimize, learn continuously, and coordinate and collaborate. It unifies the data-driven and model-driven approaches at the core of AI and Operations Research (OR). Its methodology thrusts include a new generation of hybrid optimization solvers that learn to optimize, end-to-end learning and optimization to tightly integrate forecasting and decision making, and novel machine-learning methods based on combinatorial optimization. To learn and optimize at massive scales, the Institute will contribute innovations in compact representations, data compression, and probabilistic modeling. To enable safe and scalable decision-making in uncertain and multi-agent environments that often arise in engineering disciplines, the Institute will design new methods in reinforcement learning, decentralized optimization, and data-driven optimization at massive scales. Importantly, to ensure that these scientific advances serve the interests of society, a transversal thrust will integrate ethics and values in complex systems design from inception through design and operation.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.