JZJunhui Zhang

Junhui Zhang

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Junhui Zhang has a total of 11 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Robust environmental exposure pattern recognition and outlier detection with the open-source R package pcpr

    by Lawrence G. Chillrud, Jaime Benavides, Elizabeth A. Gibson, Junhui Zhang, Jingkai Yan, John N. Wright, Jeff Goldsmith, Marianthi-Anna Kioumourtzoglou

    Background: Pattern recognition in high-dimensional mixture data is of increasing interest in environmental health (EH), as researchers often aim to identify sources or behaviors leading to potentially harmful exposures. Principal component pursuit (PCP)—a robust dimensionality reduction technique—has been successfully utilized for pattern recognition in a number of EH studies. PCP decomposes an exposure matrix into a low-rank matrix encoding consistent exposure patterns and a sparse matrix isolating outlying exposure events. However, PCP’s application has been hindered by a lack of available software tailored specifically to EH research. Methods: We introduce an open-source R package, pcpr, enabling easy PCP deployment in EH research. The package provides functions to fit and fine-tune PCP models with three EH-specific extensions: (1) a non-negativity constraint on the low-rank matrix, enhancing interpretability; (2) procedures to accommodate missingness; and (3) a specialized penalty for observations below the analytic limit of detection. We illustrate core functionality by conducting a fully reproducible source-apportionment analysis of an air pollution mixture of 26 PM 2.5 constituents measured every few days in Queens, New York City (2015–2021). Results: The functions in pcpr autonomously extracted four consistent exposure patterns (secondary signal and tailpipe emissions, traffic, crustal dust, and salt) and 211 outlying exposure events (including July Fourth firework-related spikes) from the Queens PM 2.5 mixture. If found to be associated with adverse health outcomes in downstream health models, these exposure patterns and extreme events can support targeted regulatory action. Conclusion: The pcpr package facilitates robust, reproducible, and accessible exposure pattern recognition tailored to environmental epidemiology.

  • Quadratic Memory Is Necessary for Optimal Query Complexity in Convex Optimization: Center of Mass Is Pareto Optimal

    by Moïse Blanchard, Junhui Zhang, Patrick Jaillet

    We give query complexity lower bounds for convex optimization and the related feasibility problem. We show that quadratic memory is necessary to achieve the optimal oracle complexity for first-order convex optimization. In particular, this shows that center-of-mass cutting-plane algorithms in dimension d, which use [ ] memory and [ ] queries, are Pareto optimal for both convex optimization and the feasibility problem, up to logarithmic factors. Precisely, building upon techniques introduced in previous works, we prove that to minimize 1-Lipschitz convex functions over the unit ball to [ ] accuracy, any deterministic first-order algorithms using at most [ ] bits of memory must make [ ] queries for any [ ]. For the feasibility problem, in which an algorithm only has access to a separation oracle, we show a stronger trade-off; for at most [ ] memory, the number of queries required is [ ]. This resolves a Conference on Learning Theory 2019 open problem. Funding: This work was partly supported by the Air Force Office of Scientific Research [Grant FA9550-19-1-0263] and the Office of Naval Research [Grant N00014-18-1-2122].


  • Distributionally Constrained Black-Box Stochastic Gradient Estimation and Optimization

    by Henry Lam, Junhui Zhang

    Constructing stochastic gradient estimators and descent algorithms is a core problem in stochastic optimization. In the black-box setting, namely when only noisy function evaluations are available, such gradient estimators are typically constructed via finite-differencing. In this regard, there have been schemes that aim to obtain gradient estimators for potentially many dimensions simultaneously via only few sample observations or simulation runs. However, for problems with probability simplex constraints, which arise in a range of applications from distributionally robust analysis to inverse model calibration, these schemes run into challenges one way or another when attempting to balance bias-variance in a constraint-compatible manner. This motivates us to create a new design framework for random perturbation generators and estimation schemes that bypass these challenges. Our culminated class of estimators, which is based on the Dirichlet mixtures, is demonstrably effective in distributionally constrained gradient estimation and optimization under various black-box settings.


  • pcpr: Principal Component Pursuit for Environmental Epidemiology

    by Lawrence G. Chillrud, Jaime Benavides, Elizabeth A. Gibson, Junhui Zhang, Jingkai Yan, John N. Wright, Jeff Goldsmith, Marianthi-Anna Kioumourtzoglou

    Implementation of the pattern recognition technique Principal Component Pursuit tailored to environmental health data, as described in Gibson et al (2022) .