Sungho Shin has been named the 2023 W. David Smith, Jr. Graduate Publication Award winner. for the paper “Exponential decay of sensitivity in graph-structured nonlinear programs,” which appeared in the SIAM Journal of Optimization in 2022.
This award is given to an individual for outstanding published work in chemical engineering computing and systems technology based on work done by the individual while pursuing graduate or undergraduate studies.
Shin received his Ph.D. in chemical engineering from the University of Wisconsin-Madison and has been a postdoctoral appointee in the Mathematics and Computer Science Division at Argonne National Laboratory for the past two years. His research was initially funded by the MACSER project, and he has been working with MACSER researchers since 2018 on problems including large-scale nonlinear optimization and control theory and their applications to energy systems.
In the award-winning paper, Shin and his MACSER colleagues examined the solution sensitivity of graph-structured nonlinear programs (NLPs). Such problems have many applications, including dynamic optimization, multistage stochastic programs, network optimization, and optimization with partial differential equations. The researchers developed a unifying abstraction for structured problems that they then used to establish the fundamental properties of these seemingly different problem classes.
For example, they showed that given two nodes, the solution sensitivity of one node decays exponentially as the distance from the point of data perturbation at the other node increases. Moreover, they presented conditions under which the decay rate remains uniformly bounded, allowing them to characterize the sensitivity behavior of NLPs defined over subgraphs of infinite graphs.
“Our work carries substantial implications for designing scalable computational frameworks for complex decision-making problems,” Shin said. “The findings offer valuable insights into leveraging the underlying system structure to craft effective decentralization and decomposition strategies.”
For the full paper, see Sungho Shin, Mihai Anitescu, and Victor M. Zavala, Exponential decay of sensitivity in graph-structured nonlinear programs, SIAM J. Optimization 32 (2) https://doi.org/10.1137/20M1381691