AF: Small: Is the Simulation of Quantum Many-Body Systems Feasible on the Cloud?

模拟具有叠加、干涉和纠缠等独特效应的量子力学是包括超级计算机和具有大量服务器的计算机云在内的经典计算系统的难题。要想利用计算机云有效地模拟大型量子力学系统，必须克服一些主要障碍。本文研究了凝聚态物理研究中出现的收缩张量网络的最优算法。这些算法最大限度地减少了计算机云节点之间所需的通信，并利用了其分层组织。这项工作的更广泛的研究目标是为大数据应用程序最佳地利用分层组织系统的体系结构，这些应用程序表现出细粒度的并行性。本研究项目旨在寻找新的有效方法来模拟大量子系统，这对量子信息处理、凝聚态物理、材料科学和化学都很重要。它的目标是设计和实现新的并行和分布式模拟算法，优化云计算环境，如亚马逊网络服务和国家科学基金会？科学计算的未来云。该项目的最终动机是使世界范围内的研究人员能够在合理的时间和合理的预算内可靠地模拟量子系统的大小，从而大大突破量子系统的界限。虽然该研究主要集中在凝聚态系统特性研究的有效算法及其实现上，但它也试图为其他类别的应用（例如人工智能和机器学习）推导通用策略。

Simulating quantum mechanics with its unique effects such as superposition, interference, and entanglement is a hard problem for classical computing systems including supercomputers and computer clouds with a very large number of servers. To efficiently simulate large quantum-mechanical systems using a computer cloud one has to overcome major obstacles. This research investigates optimal algorithms for contracting tensor networks which arising in the study of condensed matter physics. These algorithms minimize the required communication between the nodes of the computer cloud and exploit its hierarchical organization. The broader research objective in this effort is to optimally exploit the architecture of hierarchically organized systems for big data applications that exhibit fine-grained parallelism.This research project aims to find new efficient methods for simulating large quantum systems that are important for quantum information processing, condensed matter physics, materials science, and chemistry. Its goals are to design and implement novel parallel and distributed simulation algorithms optimized for cloud computing environments such as Amazon Web Services and the National Science Foundation?s future cloud for scientific computing. The ultimate motivation of this project is to enable researchers world-wide to significantly push the boundary in terms of the size of quantum systems that they can simulate reliably and within a reasonable time and with a reasonable budget. While the research mainly concentrates on efficient algorithms and their implementation for the study of properties of condensed matter systems, it also attempts to derive generic strategies to other classes of applications, for instance, in artificial intelligence and in machine learning.