SLSu Li

Su Li

*This is an algorithmically generated author record based on publicly available information.

Su Li has a total of 7 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • A Convex Relaxation-Based Spatial Branching Approach for Optimal Robust Group Testing Designs Under Prevalence Rate and Dilution Behavior Uncertainty

    by Su Li, Hrayer Aprahamian, Sohom Chatterjee

    Group testing is a widely adopted strategy for screening large populations for infectious diseases. Its efficiency is heavily influenced by the prevalence rate and the dilution effect of pooling, a phenomenon in which test accuracy deteriorates for large group sizes. Both factors are highly uncertain, motivating the need for robust testing schemes. In this paper, we introduce a novel regret-based formulation of the Dorfman group testing problem that accounts for uncertainty in both the prevalence rate and the dilution behavior of the assay. To solve the resulting minimax regret problem, we recast it as a more tractable conventional minimax problem and solve the nonconvex reformulation via spatial branching with convex relaxations. We derive theoretical properties for efficiently constructing convex underestimators that are guaranteed to converge to the original objective and use these to prove the algorithm’s convergence to an [ ]-optimal solution. A case study on COVID-19 with real clinical data demonstrates the algorithm’s efficiency and shows that robust testing schemes reduce maximum regret while improving both testing costs and classification accuracy. History: Accepted by J. Paul Brooks, Area Editor for Applications in Biology, Medicine, & Healthcare. Funding: This material is based on work supported in part by the National Science Foundation [Grant 2414715]. Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information ( https://pubsonline.informs.org/doi/suppl/10.1287/ijoc.2023.0465 ) as well as from the IJOC GitHub software repository ( https://github.com/INFORMSJoC/2023.0465 ). The complete IJOC Software and Data Repository is available at https://informsjoc.github.io/ .

  • Code and Data Repository for A Convex Relaxation-based Spatial Branching Approach for Optimal Robust Group Testing Designs under Prevalence Rate and Dilution Behavior Uncertainty

    by Su Li, Hrayer Aprahamian, Sohom Chatterjee

    The software and data in this repository are a snapshot of the software and data that were used in the research reported on in the paper A Convex Relaxation-based Spatial Branching Approach for Optimal Robust Group Testing Designs under Prevalence Rate and Dilution Behavior Uncertainty by S. Li, H. Aprahamian, and S. Chatterjee.


  • An Optimization-Based Order-and-Cut Approach for Fair Clustering of Data Sets

    by Su Li, Hrayer Aprahamian, Maher Nouiehed, Hadi El-Amine

    Machine learning algorithms have been increasingly integrated into applications that significantly affect human lives. This surged an interest in designing algorithms that train machine learning models to minimize training error and imposing a certain level of fairness. In this paper, we consider the problem of fair clustering of data sets. In particular, given a set of items each associated with a vector of nonsensitive attribute values and a categorical sensitive attribute (e.g., gender, race, etc.), our goal is to find a clustering of the items that minimizes the loss (i.e., clustering objective) function and imposes fairness measured by Rényi correlation. We propose an efficient and scalable in-processing algorithm, driven by findings from the field of combinatorial optimization, that heuristically solves the underlying optimization problem and allows for regulating the trade-off between clustering quality and fairness. The approach does not restrict the analysis to a specific loss function, but instead considers a more general form that satisfies certain desirable properties. This broadens the scope of the algorithm’s applicability. We demonstrate the effectiveness of the algorithm for the specific case of k-means clustering as it is one of the most extensively studied and widely adopted clustering schemes. Our numerical experiments reveal the proposed algorithm significantly outperforms existing methods by providing a more effective mechanism to regulate the trade-off between loss and fairness. History: Rema Padman served as the senior editor for this article. Data Ethics & Reproducibility Note: The code capsule is available on Code Ocean at https://doi.org/10.24433/CO.9556728.v1 and in the e-Companion to this article (available at https://doi.org/10.1287/ijds.2022.0005 ).

  • Spatiotemporal vaccine allocation policies for epidemics with behavioral feedback dynamics

    by Julius Barth, Su Li, Hrayer Aprahamian, Diwakar Gupta

    Motivated by the COVID‐19 pandemic, we study how a public health authority may allocate vaccines from a limited stockpile to different jurisdictions over time. We propose an epidemiological model with time‐varying contact rates determined by a stylized behavioral feedback mechanism to reflect multi‐wave transmission dynamics. We evaluate the performance of various information‐sensitive allocation policies (e.g., allocation proportional to local incidence) as alternatives to the widely used pro‐rata policy. We also obtain optimized allocation strategies under the proposed epidemiological model with fairness and implementable freeze‐period constraints. For the case of a multi‐wave epidemic as represented by our compartmental model with behavioral feedback, we find that none of the alternative policies offers consistently more efficient allocations than a simple pro‐rata policy across a broad range of behavioral parameter settings. In fact, in some cases the alternative policies may actually result in less efficient allocations than the pro‐rata policy. Thus our results support the conclusion that the widely used pro‐rata policy can be well justified because it is simple to explain/implement and does not cause unexpected adverse effects. However, if policy makers are willing to invest in more tailored strategies based on numerical optimization, then the identified optimized strategies are a more favorable option as they allow for a more efficient allocation of vaccines.


  • Quantifying the benefits of customized vaccination strategies: A network‐based optimization approach

    by Su Li, Hrayer Aprahamian

    We study the problem of designing vaccine distribution strategies that maximally mitigate the negative impact of an infectious disease outbreak. This is achieved through a multiperiod optimization‐based framework that embeds important subject‐specific risk and contact information into the decision‐making process. By analyzing the structure of the resulting optimization problem, we identify key structural properties which we use to construct a globally convergent solution scheme (suitable for smaller problem instances) and two, more scalable, heuristic schemes. We demonstrate the benefits of the considered framework through a case study on COVID‐19 in Texas. Our results highlight the importance of considering risk and contact information as doing so substantially reduces the total expected number of fatalities over conventional compartmental‐based approaches. These findings indicate that customization can have a significant benefit, particularly for community‐scale planning.


  • An optimization-based framework to minimize the spread of diseases in social networks with heterogeneous nodes

    by Su Li, Hrayer Aprahamian

    We provide an optimization-based framework that identifies social separation policies to mitigate the spread of diseases in social networks. The study considers subject-specific risk information, social structure, and the negative economic impact of imposing restrictions. We first analyze a simplified variation of the problem consisting of a single period and a specific social structure to establish key structural properties and construct a tailored globally-convergent solution scheme. We extend this solution scheme to heuristically solve the more general model with multiple time periods and any social structure. We use real COVID-19 data to illustrate the benefits of proposed framework. Our results reveal that the optimized policies substantially reduce the spread of the disease when compared to existing benchmark algorithms and policies that are based on a single risk factor. In addition, we utilize the considered framework to identify important subject attributes when distributing Personal Protective Equipment (PPE). Moreover, results reveal that the optimized policies continue to outperform under a more realistic setting. Our results underscore the importance of considering subject-specific information when designing policies and provide high-level data-driven observations to policy-makers that are tailored to the specific risk profile of the population that is being served.