Lune

Profiles2020-2026

Murat A Erdogdu

University of Toronto and Vector Institute

23Papers
16Last author
22Unique co-authors
0First author
0Solo

First and last author come from byline position, which some venues assign alphabetically.

Recent papers23

  1. Post-Training with Policy Gradients: Optimality and the Base Model BarrierAuthor 2 of 2ICML 2026 · 1 citation
  2. Learning quadratic neural networks in high dimensions: SGD dynamics and scaling lawsAuthor 3 of 5NeurIPS 2025 · 23 citations
  3. When Do Transformers Outperform Feedforward and Recurrent Networks? A Statistical PerspectiveAuthor 4 of 4NeurIPS 2025 · 6 citations
  4. A Geometric Analysis of PCAAuthor 2 of 3NeurIPS 2025 · 2 citations
  5. From Information to Generative Exponent: Learning Rate Induces Phase Transitions in SGDAuthor 4 of 4NeurIPS 2025 · 2 citations
  6. Learning Multi-Index Models with Neural Networks via Mean-Field Langevin DynamicsAuthor 3 of 3ICLR 2025
  7. Robust Feature Learning for Multi-Index Models in High DimensionsAuthor 3 of 3ICLR 2025
  8. A Separation in Heavy-Tailed Sampling: Gaussian vs. Stable Oracles for Proximal SamplersAuthor 4 of 4NeurIPS 2024 · 5 citations
  9. On the Efficiency of ERM in Feature LearningAuthor 3 of 3NeurIPS 2024 · 1 citation
  10. Learning in the Presence of Low-dimensional Structure: A Spiked Random Matrix PerspectiveAuthor 2 of 6NeurIPS 2023 · 47 citations
  11. Gradient-Based Feature Learning under Structured DataAuthor 4 of 4NeurIPS 2023 · 36 citations
  12. Neural Networks Efficiently Learn Low-Dimensional Representations with SGDAuthor 5 of 5ICLR 2023 · 6 citations
  13. Optimal Excess Risk Bounds for Empirical Risk Minimization on p-Norm Linear RegressionAuthor 2 of 2NeurIPS 2023 · 2 citations
  14. High-dimensional Asymptotics of Feature Learning: How One Gradient Step Improves the RepresentationAuthor 2 of 6NeurIPS 2022 · 173 citations
  15. Generalization Bounds for Stochastic Gradient Descent via Localized ε\varepsilon-CoversAuthor 3 of 3NeurIPS 2022 · 13 citations
  16. Convergence and Optimality of Policy Gradient Methods in Weakly Smooth SettingsAuthor 2 of 3AAAI 2022 · 6 citations
  17. An Analysis of Constant Step Size SGD in the Non-convex Regime: Asymptotic Normality and BiasAuthor 4 of 4NeurIPS 2021 · 66 citations
  18. Convergence Rates of Stochastic Gradient Descent under Infinite Noise VarianceAuthor 5 of 5NeurIPS 2021 · 57 citations
  19. Heavy Tails in SGD and Compressibility of Overparametrized Neural NetworksAuthor 3 of 3NeurIPS 2021 · 57 citations
  20. Fractal Structure and Generalization Properties of Stochastic Optimization AlgorithmsAuthor 3 of 6NeurIPS 2021 · 34 citations
  21. On Empirical Risk Minimization with Dependent and Heavy-Tailed DataAuthor 3 of 3NeurIPS 2021 · 22 citations
  22. Generalization of Two-layer Neural Networks: An Asymptotic ViewpointAuthor 2 of 5ICLR 2020 · 77 citations
  23. On the Ergodicity, Bias and Asymptotic Normality of Randomized Midpoint Sampling MethodAuthor 3 of 3NeurIPS 2020 · 40 citations
Dusk over the sea between two cliffs drawn in fine vertical lines