Elements: RAD Discoveries for Fundamental Physics

该项目支持一个由物理学家和计算机科学家组成的多元化团队，利用人工智能的最新进展和新的粒子探测器功能，开发新的计算和节能技术，能够以前所未有的智能检测数据中罕见和不可预测的模式。这种技术在工业、医学、国家安全以及基础物理研究中具有广泛的应用性。后者是该项目的一个重点，其中开发的基础设施可以在国家和国际粒子加速器设施中使用最先进的粒子探测器对尚未发现的物理现象进行新的搜索。这些搜索代表了科学家们几十年来在粒子物理实验中发现新现象的方法的重大偏离，因为它们不再局限于对宇宙如何运作的现有和有限的理解，也不再局限于对数据中罕见和不可预测的模式可能表现出来的预先预期。该项目旨在通过开发用于实时异常检测（RAD）的新型网络基础设施（CI），改变粒子和天体粒子物理学的基础发现。该项目旨在通过开发用于实时异常检测（RAD）的新型网络基础设施（CI），改变粒子和天体粒子物理学的基础发现。新开发的CI可以通过罕见的，不可预测的现象来发现“未知”的物理学。传统上，通过粒子和天体粒子实验发现罕见过程依赖于科学家准确预测新物理现象的能力，然后通过使用预测未知数训练的算法在数据中选择性地寻找它们。在科学发现的新时代，在前所未有的数据统计和人工智能（AI）进步的推动下，科学家现在可以使用AI驱动的工具，让数据引导期望选择性地识别可能存在于数据本身中的罕见和不可预测的签名，这些签名可能是自然界中新的基本物理现象的签名。该项目汇集了哥伦比亚和普林斯顿的物理学家和计算机科学家，包括学生，博士后研究人员和CI专业人员，以及其他美国机构和国家实验室的合作者，以开发可在资源受限的平台上执行的异常检测算法，通常用作粒子物理实验中的检测器数据处理设备。通过在现有粒子探测器设施上进行有针对性的开发和演示，该项目的最终目标是开发先进的CI，该CI具有节能，可靠，可扩展和可扩展性，因此适用于当前和未来一代粒子物理学强度，能量和宇宙前沿实验的触发系统。通过设计，交付的CI能够实现超越当前触发框架设想的基本发现，并有可能彻底改变粒子和天体粒子物理学发现的方式。开发的CI有望广泛适用于其他高吞吐量的数据选择应用程序，需要实时，智能的数据处理的目的，发现数据中的异常。这些应用可以从神经科学，医学成像，卫星成像，网络安全，该奖项由高级网络基础设施办公室颁发，由数学和物理科学理事会物理学部信息前沿物理学联合支持。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进行评估，更广泛的影响审查标准。

This project supports a diverse team of physicists and computer scientists to leverage recent advancements in AI and newly available particle detector capabilities in order to develop new computationally- and energy-efficient technology that is capable of detecting rare and unpredictable patterns in data with unprecedented intelligence. Such technology finds wide applicability in industry, medicine, national security, but also fundamental physics research. The latter is a key focus of the project, where the developed infrastructure enables new searches for yet-undiscovered physics phenomena using state-of-the-art particle detectors at national and international particle accelerator facilities. These searches represent a major departure from the methods that scientists have been applying for decades in particle physics experiments to discover new phenomena, in that they are no longer confined within an existing and limited understanding of how the Universe works, or a prior expectation for how rare and unpredictable patterns in data may manifest. The developed technology allows the data itself, rather than prior knowledge, to guide new discoveries about how Nature works at its most fundamental level, and aims to revolutionize the way new discoveries are made.This project aims to transform fundamental discoveries in particle and astro-particle physics by developing new Cyberinfrastructure (CI) for real-time anomaly detection (“RAD”). The newly developed CI enables searches that can uncover “unknown” physics through rare, unpredictable phenomena. Traditionally, discoveries of rare processes through particle and astro-particle experiments have relied on scientists’ ability to accurately predict new physics phenomena, and to subsequently selectively look for them within the data by using algorithms trained on predicted unknowns. In a new era of scientific discovery, driven by unprecedented data statistics and Artificial Intelligence (AI) advances, scientists instead may now use AI-powered tools that let the data guide expectation to selectively identify rare and unpredictable signatures that may lie within the data itself, and which may be signatures of new fundamental physics phenomena in nature. The project brings together physicists and computer scientists at Columbia and Princeton, including students, post-doctoral researchers, and CI professionals, and additional collaborators at other US institutions and national labs, to develop anomaly detection algorithms that are executable on resource constrained platforms, commonly employed as on-detector data processing devices at particle physics experiments. Through targeted developments and demonstrations at existing particle detector facilities, the ultimate goal of this project is to develop advanced CI that is energy-efficient, reliable, scalable, and expandable, and thus applicable to trigger systems of current and future generation experiments in the intensity, energy, and cosmic frontiers of particle physics. By design, the delivered CI is capable of fundamental discoveries that are beyond what is envisioned with current trigger frameworks, and has the potential of revolutionizing the way particle and astro-particle physics discoveries are made. The developed CI promises to be broadly applicable to other high-throughput data-selection applications that require real-time, intelligent data processing for the purposes of discovering anomalies in data. These applications can range from neuroscience, to medical imaging, to satellite imaging, cybersecurity, etc.This award by the Office of Advanced Cyberinfrastructure is jointly supported by the Physics at the Information Frontier in the Division of Physics within the Directorate for Mathematical and Physical Sciences.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.