Publications and Patents

Journal Articles

• Optimizing Objective Functions from Trained ReLU Neural Networks * † [pdf]
  Co-author: G. Perakis.
  Under Major Revision at Management Science. 2022 INFORMS RMP Spotlight Paper.

• A Granular View of the Emergency Department Length of Stay: Improving Predictive Power and Extracting Actionable Insights * † [pdf]
  Co-authors: M. Canellas, D. Pachamanova, G. Perakis, O. Skali Lami.
  Annals of Emergency Medicine.

• A Granular Approach to Optimal and Fair Patient Placement in Hospital Emergency Departments * † [pdf]
  Co-authors: M. Canellas, D. Pachamanova, G. Perakis, O. Skali Lami.
  Production and Operations Management (POM).

• COVID-19: Prediction, prevalence, and the operations of vaccine allocation * [pdf]
  Co-authors: G. Perakis et al.
  Manufacturing and Service Operations Management (MSOM).

• Discrete Gradient Flow Approximations of High Dimensional Evolution Partial Differential Equations via Deep Neural Networks * † [pdf]
  Co-authors: E. Georgoulis, M. Loulakis.
  Communications in Nonlinear Science and Numerical Simulation.

• The United States COVID-19 Forecast Hub Dataset [pdf]
  Co-authors: E. Cramer et al.
  Scientific Data - Nature.

• Evaluation of individual and ensemble probabilistic forecasts of COVID-19 mortality in the US [pdf]
  Co-authors: E. Cramer et al.
  Proceedings of the National Academy of Sciences (PNAS).

Conference Proceedings

• Causal LLM Routing: End-to-End Regret Minimization from Observational Data † [pdf]
  Co-authors: W. Sun, G. Perakis.
  Conference on Neural Information Processing Systems (NeurIPS) 2025.

• Overcoming the Optimizer’s Curse: Obtaining Realistic Prescriptions from ReLU Neural Networks † [pdf]
  Co-author: G. Perakis.
  International Conference on Machine Learning (ICML) 2024.

• Learning the Optimal Reconciliation for Hierarchical Time Series † [pdf]
  Co-authors: G. Perakis, W. Sun, P. Y. Chen, Y. Zhu.
  International Conference on Machine Learning (ICML) 2024.

• Manifold-Aligned Counterfactual Explanations for Neural Networks † [pdf]
  Co-authors: W. Sun, G. Perakis.
  International Conference on Artificial Intelligence and Statistics (AISTATS) 2024.

• Neural-Informed Decision Trees * † [pdf]
  Co-author: G. Perakis.
  ACM SIGKDD 2023. Workshop on Ethical Artificial Intelligence.

• Learning Prescriptive ReLU Networks * † [pdf]
  Co-author: W. Sun.
  International Conference on Machine Learning (ICML) 2023.

• A Mechanism Design and Learning Approach for Revenue Maximization on Cloud Dynamic Spot Markets † [pdf]
  Co-authors: C. Bitsakos, I. Konstantinou, D. Fotakis, N. Koziris.
  IEEE International Conference on Cloud Computing 2021.

• The Power of Analytics in Epidemiology for COVID-19: Prediction, Prevalence and Vaccine Allocation * [pdf]
  Co-authors: G. Perakis et al.
  INFORMS Conference on Service Science 2021.

• BigOptiBase: Big Data Analytics for Base Station Energy Consumption Optimization [pdf]
  Co-authors: E. Kassela, N. Provatas, I. Konstantinou, N. Koziris.
  IEEE International Conference on Big Data 2019.

Working Papers

• Generalisation Error Control for Physics Informed Neural Networks with Probabilistic Training * † [preprint out soon]
  Co-authors: C. Makridakis, M. Loulakis.

US Patents

• Counterfactual Prediction and Interpretable Policy Learning from Observational Data using Prescriptive ReLU Networks
  Co-authors: M. Ettl, W. Sun.
  US Patent App. 18/217091. Status: Published. Publication No: 20250005347.

• Manifold-Aligned Counterfactual Explanations
  Co-authors: M. Ettl, W. Sun.
  US Patent App. 18/618596. Status: Published. Publication No: 20250307599.

• Hierarchical Time Series Forecasting
  Co-authors: G. Perakis, P. Y. Chen, W. Sun, W. Y. Zhu.
  US Patent App. 18/984896. Status: Filed.

• An Integrated System for LLM Routing with Observational Data
  Co-authors: W. Sun, G. Perakis.
  Under preparation. P202501674US01 (IBMP1446US).

* names in alphabetical order
† lead author