Yuezhu (Ruby) Xu

Purdue University

Ph.D. Candidate, Industrial Engineering (Operations Research)
January 2023–May 2027 (expected) · GPA: 4.00/4.00

Dissertation committee: Sivaranjani Seetharaman (Advisor), Andrew (Lu) Liu, Liwei Jiang, Jan Drgona (JHU).

Columbia University

M.S. in Data Science
September 2021–December 2022 · GPA: 4.07/4.33

Shanghai Jiao Tong University

B.S. in Mathematics and Applied Mathematics
September 2017–June 2021 · GPA: 3.62/4.30

• Learning-enabled control and autonomous systems with provable safety and stability guarantees.

• Scalable reachability, verification, and sensitivity analysis for neural and nonlinear systems.

• Scientific machine learning, operator learning, and structure-exploiting computational methods.

Peer-Reviewed Publications

• Xu, Y. & Sivaranjani, S., ECLipsE-Gen-Local: Efficient Compositional Local Lipschitz Estimates for Deep Neural Networks. Transactions on Machine Learning Research (TMLR), 2026

• Xu, Y., Serry, M., Liu, J., & Sivaranjani, S., Learning Neural Network Safe Tracking Controllers from Backward Reachable Sets. Accept with Minor Revision, European Journal of Control, 2026

• Sivaranjani, S., Xu, Y., et al., Control-Oriented System Identification: Classical, Learning, and Physics-Informed Approaches. Annual Reviews in Control, 2026

• Xu, Y., Sivaranjani, S., & Gupta, V., Learning Neural Koopman Operators with Dissipativity Guarantees. IEEE Conference on Decision and Control (CDC), 2025

• Xu, Y. & Sivaranjani, S., ECLipsE: Efficient Compositional Lipschitz Constant Estimation for Deep Neural Networks. Accepted as spotlight paper in Advances in Neural Information Processing Systems (NeurIPS), 2024. (Top ~2% out of 15,000+ submissions in NeurIPS 2024)

• Xu, Y. & Sivaranjani, S., Learning Dissipative Neural Dynamical Systems. IEEE Control Systems Letters (LCSS), 2023. (Jointly selected for presentation in the ACC 2023)

Working Papers / Manuscripts

• Xu, Y., Sivaranjani, S., CLEAR: Certified Local Ellipsoidal Analysis of Reachability for Neural Network Control. Under Review at IEEE Control Systems Letters (L-CSS), with ACC 2027 option, 2026

• Jang, I., Xu, Y., Sivaranjani, S., Bridgeman, LJ., Neural Network Controllers for Euler-Lagrange Systems. Under Review at IEEE Control Systems Letters (L-CSS), with ACC 2027 option, 2026

• Tang, W., Xu, Y., Sivaranjani, S., MANIFEST: Manifold-Informed Learning of Feasibility-Seeking Neural Solvers for Constrained Optimization. Submitted to the International Conference on Learning Representations (ICLR), 2027

• Xu, Y. & Sivaranjani, S., Scalable and Accurate Local Lipschitz Estimation of Residual Neural Networks. Under review at the AAAI Conference on Artificial Intelligence (AAAI), 2027

• Kuang, S., Xu, Y., Sivaranjani, S., & Lin, X., Demystifying Lipschitz verification: positive matrices, negative results. arXiv preprint, 2026

• Kuang, S., Xu, Y., Sivaranjani, S., & Lin, X., Lipschitz verification of neural networks through training. arXiv preprint, 2026

• Xu, Y., Kuang, S., Lin, X, & Sivaranjani, S., ECLipsE-Fast+: Fast and Scalable SDP-Free Lipschitz Estimation. Manuscript in preparation, 2026

• Xu, Y., He, H., Sivaranjani, S., From Scalar Bounds to Ellipsoidal Geometry: Scalable Sensitivity and Reachability Analysis for Neural Networks. Manuscript in preparation, 2026

Scalable Neural Certification

• Developed scalable Lipschitz estimation methods for broad neural-network architectures, emphasizing computational efficiency and applicability across diverse model structures.

• Developed reachability and sensitivity analysis methods for neural networks and learning-enabled dynamical systems to support robustness and safety assessment.

• Studied geometric representations of uncertainty for richer and more accurate certificates.

• Built research software for robustness, sensitivity, and reachability analysis.

Scientific and Physics-Informed Machine Learning

• Developed learning-based modeling and control methods that incorporate physical structure and control-theoretic properties to provide stability, dissipativity, safety, and performance guarantees for nonlinear dynamical systems.

• Studied control-oriented system identification across classical, learning-based, and physics-informed approaches, emphasizing model structure, prior knowledge, and suitability for downstream control, and contributed a comprehensive survey.

• Dr. Theodore J. and Isabel M. Williams Fellowship in Industrial Control Systems (Graduate Fellowship), Edwardson School of Industrial Engineering, Purdue University, 2026–2027

• Audience Choice Award (Gold), Poster Session, ICON Student Research Conference 2026

• Student Travel Award, IEEE Conference on Decision and Control 2025 (Workshop & Main Conference)

• NSF Travel Award, Learning for Dynamics and Control Conference (L4DC), 2025

• NeurIPS Scholar Award, Advances in Neural Information Processing Systems (NeurIPS), 2024

• Purdue Graduate Student Government (PGSG) Travel Grant, Advances in Neural Information Processing Systems (NeurIPS), 2024

• Second Prize, Shanghai Division, China Undergraduate Mathematical Contest in Modeling, 2020

• Shanghai Jiao Tong University Undergraduate Outstanding Scholarship, Class B, 2020, 2021

• Shanghai Jiao Tong University Undergraduate Outstanding Scholarship, Class C, 2018, 2019

• 2026 — Reliable Learning-Based Decision and Control Systems: Optimization, Certification, and Robust Synthesis. Job talk in the session “Robust Optimization, Testing, and Control,” INFORMS Annual Meetings.

• 2026 — Learning Dynamical Systems with Guarantees. Contributed talk, SIAM Great Lakes Section Annual Meeting.

• 2026 — Robustness in Learning-Based Models. Edwardson Summer Seminar Series.

• 2026 — ECLipsE-NN: A Scalable Certification Framework for Neural Networks. Poster, Midwest Machine Learning Symposium (MMLS).

• 2026 — ECLipsE Framework: Efficient Compositional (Local) Lipschitz Constant Estimation for Deep Neural Networks. Lightning talk and Audience Choice Award (Gold) poster, 4th ICON Student Research Conference.

• 2025 — Learning Neural Koopman Operators with Dissipativity Guarantees. Talk in the invited session “Physics-aware Learning for Planning and Control,” 64th IEEE Conference on Decision and Control.

• 2025 — Learning Neural Dynamical Systems with Dissipativity Guarantees. Workshop Scientific Machine Learning: Theory, Algorithms and Applications (CCAM).

• 2025 — ECLipsE: Efficient Compositional Lipschitz Constant Estimation for Deep Neural Networks. Poster, 11th Midwest Workshop on Control and Game Theory.

• 2025 — Learning Dissipative Neural Dynamical Systems. Poster, 11th Midwest Workshop on Control and Game Theory.

• 2024 — ECLipsE: Efficient Compositional Lipschitz Constant Estimation for Deep Neural Networks. Spotlight poster, NeurIPS.

• 2024 — Learning Dissipative Neural Dynamical Systems. Talk, American Control Conference.

• 2024 — Learning Dissipative Neural Dynamical System. Talk, ICON 2nd Student Conference.

• 2024 — Learning Dissipative Neural Dynamical System. Talk, Grad WiE Network Symposium.

• 2023 — Learning Dissipative Neural Dynamical System. Poster, AFOSR Poster Session.

Additional Meetings

• 2025.06 — 7th Annual Learning for Dynamics & Control Conference (L4DC).

• 2025.06 — Talk at Edwardson Summer Seminar.

• 2023.10 — INFORMS Annual Meetings.

• Transactions on Machine Learning Research (TMLR), 2026

• AAAI Conference on Artificial Intelligence (AAAI), 2026

• Advances in Neural Information Processing Systems (NeurIPS), 2025–2026

• International Conference on Machine Learning (ICML), 2026

• European Control Conference (ECC), 2026

• IEEE Conference on Decision and Control (CDC), 2025–2026

• Learning for Dynamics and Control Conference (L4DC), 2025

• IEEE Control Systems Letters, 2024–2026

• Indian Control Conference (ICC), 2023–2026

• ECLipsE series: Lead developer and maintainer of an open-source MATLAB and Python package for scalable neural-network robustness, sensitivity, and certification analysis. GitHub repository (pip install eclipse-nn)

• Collaboration with Jan Drgona at Johns Hopkins University on integrating robustness and certification tools into PNNL’s Neuromancer, with technical tutorials and learning-based control examples.

• Open-source implementations of classical, learning-based, and physics-informed system identification methods for control-oriented modeling and analysis. GitHub repository

Teaching Assistant

IE 47400 Industrial Control Systems, Purdue University, Spring 2023

• Led recitations and tutorials on classical and modern control.

• Designed MATLAB/Simulink demonstrations using Control System Toolbox.

Graduate Research Mentor

Purdue University, Summer 2025–present

• Mentored junior graduate student Wangbo Tang on learning-based constrained optimization and manifold-aware training methods.

• Guided AC optimal power flow experiments and related model development and evaluation.

Methods: Robust, nonlinear, and learning-based control; Lyapunov and dissipativity theory; reachability and safety verification; system identification; Koopman and operator learning; Neural ODEs; scientific machine learning; semidefinite and spectral optimization; matrix inequalities.

Mathematical & Statistical Foundations: Dynamical systems, differential equations, optimization, probability, stochastic processes, linear algebra, matrix analysis, numerical methods, and statistical modeling.

Programming & Tools: Python (PyTorch, NumPy, SciPy, Pandas), MATLAB/Simulink (YALMIP, CORA, Control System Toolbox), R, SQL, LaTeX.