SHF: Small: Reconfigurability and Technology Integration of Magnetic Energy Minimization Co-Processor (MEMCoP)

这是两所大学——USF和UCLA的合作提议，他们提倡一种新的计算形式，使用圆形纳米磁铁来解决二次优化问题，比传统计算机快几个数量级。通过这项研究，可以潜在地加速广泛的应用领域，例如，在社交媒体，纠错代码，大数据应用等中寻找模式的计算成本高且难以并行化的问题。该项目计划吸引REU的本科生，而pi将通过与K-12教师和K7-K8女学生的互动参与当地社区。作为斯隆大学模范指导中心（UCEM）的一部分，首席PI将招募和留住本科生。加州大学洛杉矶分校的团队将与一群由卓越工程与多样性中心（CEED）赞助的本科生合作。该方案的核心计算主题是将二次能量最小化问题空间映射为一组相互作用的磁体，使问题变量之间的能量关系与相应磁体之间的能量关系成正比。优化是由磁体本身的弛豫物理完成的，解可以并行读出。本质上，给定一个特定的问题实例，该计划是要达到一个特定的磁布局，其松弛状态将是原始问题的解决方案。可重构性是这种高回报计算形式成功的基础（人们不打算制造特定的磁性布局来解决问题的一个实例）。提出了三种可供选择的编程解决方案，并将评估机制之间的权衡。该项目还计划探索可重构性以及可读性和系统集成方面。

This collaborative proposal been two universities - USF and UCLA - advocates a new form of computing that uses circular nanomagnets to solve quadratic optimization problems, orders of magnitude faster than that of a conventional computer. A wide range of application domains can be potentially accelerated through this research, e.g., computationally expensive and hard to parallelize problems of finding patterns in social media, error-correcting codes, big data applications etc. The project plans to engage undergraduate REU students, and the PIs will be involved with the local community through interaction with K-12 teachers and K7-K8 women students. As a part of Sloan University Center for Exemplary Mentoring (UCEM), the lead PI will recruit and retain undergraduate students. The UCLA team will work with a group of undergraduate students sponsored by Center for Excellence in Engineering and Diversity (CEED).Core computational theme of this proposal is mapping quadratic energy minimization problem spaces into a set of interacting magnets such that the energy relationship between the problem variables is proportional to that of the energies between the corresponding magnets. The optimization is accomplished by the relaxation physics of the magnets themselves and solutions can be read-out in parallel. In essence, given a specific instance of the problem, the plan is to arrive at a specific magnetic layout, the relaxed state of which will be the solution of the original problem. Reconfigurability is fundamental to the success of this form of high-payoff computing (one does not intend to fabricate a specific magnetic layout to solve one instance of a problem). Three alternative programming solutions are proposed, and trade-offs between the mechanisms will be evaluated. The project also plans to explore reconfigurability together with readability and system integration aspects.