Learning and Decision Intelligence
Background
Large-scale networks provide a natural representation of complex systems comprising thousands of interacting components, but their scale and structural complexity create significant challenges for optimization and decision-making. Many network optimization problems are formulated as mixed-integer programming (MIP) models, which are often NP-hard and become increasingly difficult to solve as the problem size grows. At the same time, high-dimensional data create challenges for developing predictive and decision-support models that are both accurate and interpretable. Learning-augmented optimization offers an opportunity to leverage information generated during the search process to improve decision-making, while causal inference provides complementary tools for uncovering relationships and improving the interpretability of data-driven models. However, effectively integrating learning with optimization and developing robust approaches that generalize across complex networked systems remain important challenges.
Objective
This research seeks to develop scalable learning-augmented optimization and safety-driven policies for deterministic and stochastic formulations of MIP models and multi-agent sequential decision-making on large-scale networks. We use benchmark problems and multiple city-scale networks to demonstrate the added value of our approaches relative to baseline optimization and reinforcement learning algorithms in terms of solution quality, convergence rate, and computational time.
Publications
- Mohebbi, S., Pamukcu, E., Bozdogan, H., (2019). “A new data adaptive elastic net predictive model using hybridized smoothed covariance estimators with information complexity”, Journal of Statistical Computation and Simulation, 89(6), 1060-1089.
- Aslani, B., Mohebbi, S., (2023). “Ensemble framework for causality learning with heterogeneous Directed Acyclic Graphs through the lens of optimization”, Computers and Operations Research, 152, 106148.
- Aslani, B., Mohebbi, S., Ougthon. E. (2024). “A systematic review of optimization methods for recovery planning in cyber-physical infrastructure networks: current state and future trends”, Computers & Industrial Engineering, 192, 110224. https://doi.org/10.1016/j.cie.2024.110224
- Aslani, B., Mohebbi, S., (2024). “Learn to decompose multi-objective optimization models for large-scale networks”, International Transactions in Operational Research, 31(2), 949-978. https://doi.org/10.1111/itor.13169
- Aslani, B., Mohebbi, S., (2025). “A learning-augmented branch-and-price for large-scale integrated network design and scheduling problem in road restoration” [Under Review; 2024 Best Track Paper Award of the IISE Operations Research Division].
- Aslani, B., Mohebbi, S., Ji, R., (2025). “Learn-to-construct cuts in nested benders decomposition with application to large-scale stochastic multi-stage network design and scheduling” [Under Review]. Available at SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5895134
- Murali, P.S., Mohebbi, S., (2026). “Safety-driven decentralized decisions in networks: An indicator-based reinforcement learning approach”, IISE Transactions, 1–13, https://doi.org/10.1080/24725854.2026.2646938
Funding Source
Mission-Focused Applied Prototyping, Air Force Research Laboratory, AFCENT (2023-2026), Co-PI, $7,442,840.
Center for Resilient and Sustainable Communities, George Mason University (2021-2023), PI, $31,640.
Postdoctoral Scholar and GRAs
Dr. Babak Aslani (Postdoc), Pavithra Sripathanallur Murali (PhD Student), Andrew Moseman (MSc student)