About

I am a third-year PhD student in the Department of Electrical Engineering at Stanford University, advised by Stephen Boyd. Prior to joining Stanford, I received a BSc in Physics and an MSc in Applied Mathematics, both from Linköping University, Sweden. My research focuses on developing software and algorithms for solving optimization problems.

Publications

D. Cederberg, S. Boyd. Presolving for GPU-Accelerated First-Order LP Solvers. arXiv preprint arXiv:2604.23951, 2026.
📄 Paper | 💻 Code |
D. Cederberg. Tyler's M-estimator Through the Lens of Convex-Concave Programming. AISTATS, 2026 (Spotlight).
📄 Paper
D. Cederberg. Fitting a Robust Factor Model via Expectation-Maximization. IEEE Transactions on Signal Processing, 2026.
📄 Paper | 💻 Code
D. Cederberg, S. Boyd. Projections onto Spectral Matrix Cones. arXiv preprint arXiv:2511.01089, 2025.
📄 Paper | 💻 Code | 📊 Slides
D. Cederberg, X. Wu, S. Boyd, M. Johansson. An Asynchronous Bundle Method for Distributed Learning Problems. International Conference on Learning Representations (ICLR), 2025.
📄 Paper
D. Cederberg, E. Larsson, M. Johansson. FABLE: a bundle method for federated le- arning in wireless systems. IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2025.
📄 Paper
D. Cederberg, First-order Methods for Nonnegative Trigonometric Matrix Polynomials. Journal of Optimization Theory and Applications, 2025.
📄 Paper | 💻 Code 📊 Slides
D. Cederberg, Toeplitz covariance estimation with applications to MUSIC. Journal of Signal Processing, 2024.
📄 Paper | 💻 Code
D. Cederberg, A. Hansson, A. Rantzer, Synthesis of Minimax Adaptive Controller for a Finite Set of Linear Systems. IEEE Conference on Decision and Control, 2022.
📄 Paper | 💻 Code

Software

During my PhD, I have developed several software packages, most of them written in C.

PSLP, a fast presolver for large-scale linear programs. PSLP is used by Nvidia in its GPU-accelerated optimization solver cuOpt, among others.
DNLP, an extension of CVXPY to (nonconvex) nonlinear programming. This includes the development of the Sparse Differentiation engine for efficiently computing sparse Jacobians and Hessians. DNLP is available in CVXPY version 1.9.0 and later.
Presumably the World's fastest algorithm for computing the maximum likelihood estimate of a Toeplitz covariance matrix.

I have also contributed extensively to the semidefinite programming functionality of the open-source convex conic solver SCS.

Teaching

In the summer of 2025 I had the privilege of being the instructor for EE 364a: Convex Optimization I at Stanford. I've also enjoyed serving as a teaching assistant for EE 364a (2024) and ENGR108 (2025) at Stanford, and for several classes and multiple offerings in statistics and optimization during my undergraduate studies at Linköping University, including TAMS65, TAOP33, TAOP07, and TAOP88.