CC* Compute: CAML - Accelerating Machine Learning via Campus and Grid

机器学习是人工智能领域的一个子集，它在一系列问题上取得了令人印象深刻的进展，从识别图像中的物体到将法语翻译成英语。在圣母大学，加速机器学习的网络基础设施（CAML）资源允许整个大学的教师利用机器学习来解决他们学科内的问题。CAML使用图形处理单元（gpu）来显著提高训练机器学习算法的速度，使研究能够更快地解决问题，并训练能够解决更困难问题的更复杂模型。CAML受益于广泛的研究活动，从在大型强子对撞机上寻找新粒子到探索导致医学突破的新化学物质。作为开放科学网格的一部分，CAML还使更广泛的社区受益，为来自美国各地大学和实验室的研究人员提供服务。此外，CAML还用于从高中到研究生院的学生的教育和推广，帮助培养下一代应对公共和私营部门的数据科学挑战。CAML通过开放科学网格（OSG）为圣母大学本地和全国的研究界提供加速机器学习的GPU资源。CAML物理托管适合加速标准深度学习库模型训练的GPU资源，但也允许按需云访问更多实验性架构，如FPGA资源。配置为交互式和批处理访问，CAML既支持小规模探索，也支持大规模科学发现。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Machine learning, a subset of the field of artificial intelligence, has enabled impressive progress on a range of problems from identifying objects in images to translating French into English. At the University of Notre Dame, the Cyberinfrastructure to Accelerate Machine Learning (CAML) resource allows faculty across the university to leverage machine learning to address problems within their disciplines. CAML uses graphical processing units (GPUs) to provide a significant boost in the speed of training machine learning algorithms, enabling researching to solve problems faster and to train more complex models capable of addressing more difficult problems. CAML benefits a wide range of research activities, from searching for new particles at the Large Hadron Collider to exploring new chemicals leading to medical breakthroughs. CAML also benefits the broader community as part of the Open Science Grid, serving researchers from universities and labs across the US. Furthermore, CAML is used for education and outreach involving students ranging from high school to graduate school, helping to train the next generation to tackle data science challenges in the public and private sector.CAML provides GPU resources for accelerating machine learning to the research community both locally at the University of Notre Dame and nationally through the Open Science Grid (OSG). CAML physically hosts GPU resources suitable for accelerating the training of models from standard Deep Learning libraries, but also enables on-demand cloud access to more experimental architectures like FPGA resources. Configured for both interactive and batch access, CAML supports both small-scale explorations to large-scale discovery science.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.