Collaborative Research: Optimized Testing Strategies for Fighting Pandemics: Fundamental Limits and Efficient Algorithms

合作研究：抗击流行病的优化测试策略：基本限制和高效算法

• 批准号: 2133170
• 负责人: Jing Yang
• 金额: $ 27.48万
• 依托单位: Pennsylvania State Univ University Park
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2022
• 资助国家: 美国
• 起止时间: 2022-09-01 至 2025-08-31
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2133170&HistoricalAwards=false
• 关键词: Collaborative; Research; Optimized; Testing; Strategies

Large-scale high-throughput prevalence and diagnostic testing is essential for the containment and mitigation of pandemics. The testing bottleneck in the COVID-19 pandemic has led to a resurgence of interest in group testing, where several people's biological samples are mixed together and examined in a single test. When the rate of infection in the population is low, this method can significantly reduce the total number of tests per subject and increase the throughput of the existing testing infrastructure. However, traditional group testing has the following limitations: First, efficient group testing based methods for the estimation of prevalence have been largely overlooked in the literature. Second, traditional group testing usually assumes that the testing results are qualitative (positive versus negative), not quantitative (providing viral load information). Third, the theoretical study of group testing rarely takes practical constraints, such as the sensitivity of the pooled tests and the dilution effect, into consideration, which hinders the applicability of the testing schemes in practice. The goal of this project is to overcome these limitations of traditional group testing and design advanced pooled testing strategies for efficient prevalence tracking and accurate infection diagnosis. It will develop optimized pooled testing strategies with strong theoretical performance guarantees yet feasible and cost-effective in practice.The proposed research is organized in three research thrusts as follows. Thrust 1 aims to design effective sampling and testing algorithms to estimate the prevalence in communities and track its evolution, under scarce testing resource constraints. Thrust 2 focuses on the design of optimized pooling and decoding algorithms for compressed sensing based (COVID-19) virus diagnostic testing. Thrust 3 validates the accuracy and efficiency of the proposed pooled testing through experiments on anonymized COVID-19 samples. This project bridges group testing and online learning, the two largely disconnected areas, with the objective to effectively allocate limited testing resources for efficient prevalence tracking. Such integration leads to novel sampling strategies, broadens the paradigm of group testing, and advances the state of the art of online learning. Moreover, the proposed compressed sensing based diagnostic testing leverages quantitative measurements provided by advanced testing technologies, which can significantly increase test throughput, reduce the number of needed tests, decrease the consumption of scarce reagents, and provide results robust against observation noises and outliers. The rich compressed sensing theory provides possible approaches to the rigorous mathematical certification of the correctness of the decoded results. Besides, the clinical constraints on pooled testing also lead to novel problem formulation and theoretical characterization, enriching the study of compressed sensing.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.

大规模的高通量流行和诊断性检测对于遏制和缓解流行病至关重要。新冠肺炎疫情中的检测瓶颈重新引发了人们对集体检测的兴趣，即几个人的生物样本混合在一起，在一次检测中进行检查。当人群感染率较低时，这种方法可以显著减少每个受试者的总检测次数，并增加现有检测基础设施的吞吐量。然而，传统的群体测试有以下局限性：首先，基于群体测试的有效的流行率估计方法在很大程度上被文献所忽视。其次，传统的群体检测通常假设检测结果是定性的(阳性与阴性)，而不是定量的(提供病毒载量信息)。第三，分组测试的理论研究很少考虑实际约束，如混合测试的敏感性和稀释效应，这阻碍了测试方案在实践中的适用性。该项目的目标是克服传统群体检测的这些局限性，并设计先进的联合检测策略，以实现有效的流行跟踪和准确的感染诊断。它将开发优化的混合测试策略，在理论上有很强的性能保证，但在实践中是可行的和具有成本效益的。推力1旨在设计有效的抽样和测试算法，在有限的测试资源限制下，估计社区中的流行率并跟踪其演变。推力2专注于为基于压缩感知(新冠肺炎)的病毒诊断测试设计优化的池化和解码算法。推力3通过对匿名新冠肺炎样本的实验验证了所提出的混合测试的准确性和效率。该项目将团体测试和在线学习这两个基本上互不相连的领域联系起来，目的是有效地分配有限的测试资源，以便有效地跟踪流行率。这种整合导致了新的抽样策略，拓宽了团体测试的范式，并促进了在线学习的艺术水平。此外，提出的基于压缩传感的诊断测试利用了先进测试技术提供的定量测量，可以显著增加测试吞吐量，减少所需测试的数量，减少稀缺试剂的消耗，并提供对观测噪声和异常值具有鲁棒性的结果。丰富的压缩感知理论为对解码结果的正确性进行严格的数学证明提供了可能。此外，联合测试的临床限制也导致了新的问题表达和理论表征，丰富了压缩感觉的研究。这一奖项反映了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。

项目成果

Federated Linear Contextual Bandits with User-level Differential Privacy

• DOI: 10.48550/arxiv.2306.05275
• 发表时间: 2023-06
• 作者: Ruiquan Huang;Huanyu Zhang;Luca Melis;Milan Shen;Meisam Hajzinia;J. Yang

Exploiting Feature Heterogeneity for Improved Generalization in Federated Multi-task Learning

• DOI: 10.1109/isit54713.2023.10206757
• 发表时间: 2023-06
• 期刊: 2023 IEEE International Symposium on Information Theory (ISIT)
• 作者: Renpu Liu;Jing Yang;Cong Shen

Provably Efficient Offline Reinforcement Learning with Perturbed Data Sources

• DOI: 10.48550/arxiv.2306.08364
• 发表时间: 2023-06
• 期刊: ArXiv
• 作者: Chengshuai Shi;Wei Xiong;Cong Shen;Jing Yang