Curriculum Vitae

Yunus Emre Demirci

Department of Mathematics, University of Michigan. Ann Arbor, MI, USA · edemirci@umich.edu. A PDF version is available here.

Education

Research Interests

My research interests lie broadly in stochastic control, nonlinear filtering, and partially observed Markov decision processes (POMDPs). I am particularly interested in the robustness, stability, and approximation of control and filtering systems under partial information.

In discrete-time settings, I have studied the impact of model mismatch and quantization on the stability and convergence of belief-state dynamics, deriving explicit bounds on value functions and identifying conditions for near-optimality. In particular, I have investigated the Wasserstein and total-variation continuity of filter kernels under model perturbations, robustness in the average-cost setting, and unique ergodicity results for control-free models.

More recently, I have been focusing on continuous-time stochastic systems under partial observation, especially diffusion processes — studying robustness to prior mismatch and how discrete-time approximations via time and observation quantization can be used to analyze and control such systems.

Publications

Preprints

  1. Demirci, Y. E., Pradhan, S.., Yüksel, S. Near Optimal POMDP Approximation of Controlled Diffusions with Partial Information.
  2. Demirci, Y. E., Kara, A. D., Yüksel, S. Sensitivity of Filter Kernels and Robustness to Incorrect Transition and Measurement Kernel Perturbations in Partially Observable Stochastic Control. arXiv:2508.10658v3.
  3. Demirci, Y. E., Kara, A. D., Yüksel, S. Refined Bounds on Near Optimality Finite Window Policies in POMDPs and Their Reinforcement Learning. arXiv:2409.04351.

Publications

  1. Demirci, Y. E., and Yüksel, S. (2025). Unique Ergodicity of Non-Linear Filters via Reachability and Uniform Weak Continuity. Electronic Communications in Probability, 30, 82, 1–15. https://doi.org/10.1214/25-ECP732
  2. Demirci, Y. E., Kara, A. D., and Yüksel, S. (2024). Average Cost Optimality of Partially Observed MDPs: Contraction of Non-Linear Filters and Existence of Optimal Solutions and Approximations. SIAM Journal on Control and Optimization, 62(6), 2859–2883. https://doi.org/10.1137/24M1643736
  3. Demirci, Y., Işlak, Ü., Özdemir, A. (2023). Mixing time bounds for edge flipping on regular graphs. Journal of Applied Probability, 1–16. https://doi.org/10.1017/jpr.2023.3
  4. Çiçeksiz, A., Demirci, Y. E., Işlak, Ü. (2023). The Variance and the Asymptotic Distribution of the Length of Longest k-alternating Subsequences. Discrete Mathematics and Theoretical Computer Science, 25(1). https://doi.org/10.46298/dmtcs.10296
  5. Demirci, Y. E., Ekim, T., Yıldız, M. A. (2023). Defective Ramsey Numbers and Defective Cocolorings in Some Subclasses of Perfect Graphs. Graphs and Combinatorics, 39(1), 18. https://doi.org/10.1007/s00373-023-02612-4
  6. Demirci, Y. E., Işlak, Ü., Yıldız, M. A. (2023). On a cover time problem on a dynamic graph with steps at random times. Statistics and Probability Letters, 197, 109821. https://doi.org/10.1016/j.spl.2023.109821
  7. Demirci, Y. E., Ekim, T., Gimbel, J., Yıldız, M. A. (2021). Exact values of defective Ramsey numbers in graph classes. Discrete Optimization, 42, 100673. https://doi.org/10.1016/j.disopt.2021.100673

Conference Proceedings

  1. Demirci, Y. E., Kara, A. D., & Yüksel, S. (2025). Sensitivity of Filter Kernels and Robustness to Incorrect Transition and Measurement Kernel Perturbations in Partially Observable Stochastic Control. In Proceedings of the 2025 IEEE 64th Conference on Decision and Control (CDC 2025), Rio de Janeiro, Brazil, December 10-12, 2025.
  2. Demirci, Y. E., Kara, A. D., & Yüksel, S. (2025). Wasserstein Regularity of Nonlinear Filters As Belief-MDPs, and Implications on Ergodicity, Optimality and Learning for POMDPs. In Proceedings of the 2025 American Control Conference (ACC 2025), Denver, Colorado, USA, July 8-10, 2025.
  3. Demirci, Y. E., Kara, A. D., & Yüksel, S. (2024). Average Cost Optimality of Partially Observed MDPs: Contraction of Non-Linear Filters and Existence of Optimal Solutions and Approximations. In Proceedings of the 2024 American Control Conference (ACC 2024), Toronto, Canada, July 8-12, 2024.

Honors & Awards

Scholarships

Teaching Experience

More on the teaching page.

Talks

Languages & Technical Skills