CV

Education

• Ph.D. in Computer Science, Linköping University, 2025 (expected)
• M.Sc. in Applied Physics and Electrical Engineering, Linköping University, 2020
  • Track: Theory, modeling, and computations
  • Thesis title: Graph neural networks for prediction of formation energies of crystals

Work experience

• Summer 2019: Machine learning intern
  • Ericsson Research, Linköping
  • Data collection and analysis

Skills

• Machine learning
  • Deep learning
  • Graph neural networks
  • Diffusion models
  • Sequential Monte Carlo
• Python programming
  • Implementations of machine learning methods

Publications

WyckoffDiff – A Generative Diffusion Model for Crystal Symmetry

Ekström Kelvinius, F., Andersson, O. B., Parackal, A. S., Qian, D., Armiento, R., & Lindsten, F. (2025). WyckoffDiff-A Generative Diffusion Model for Crystal Symmetry. International Conference on Machine Learning (to appear)

Solving Linear-Gaussian Bayesian Inverse Problems with Decoupled Diffusion Sequential Monte Carlo

Ekström Kelvinius, F., Zhao, Z., & Lindsten, F. (2025). Solving Linear-Gaussian Bayesian Inverse Problems with Decoupled Diffusion Sequential Monte Carlo. International Conference on Machine Learning (to appear)

Discriminator Guidance for Autoregressive Diffusion Models

Ekström Kelvinius, F. & Lindsten, F.. (2024). Discriminator Guidance for Autoregressive Diffusion Models. Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 238:3403-3411 Available from https://proceedings.mlr.press/v238/ekstrom-kelvinius24a.html.

Accelerating Molecular Graph Neural Networks via Knowledge Distillation

Ekström Kelvinius, F., Georgiev, D., Toshev, A., & Gasteiger, J. (2024). Accelerating molecular graph neural networks via knowledge distillation. Advances in Neural Information Processing Systems, 36.

Autoregressive Diffusion Models with non-Uniform Generation Order

Kelvinius, F. E., & Lindsten, F. (2023). Autoregressive Diffusion Models with non-Uniform Generation Order. In ICML 2023 Workshop on Structured Probabilistic Inference & Generative Modeling.

Graph-based machine learning beyond stable materials and relaxed crystal structures

Ekström Kelvinius, F., Armiento, R., & Lindsten, F. (2022). Graph-based machine learning beyond stable materials and relaxed crystal structures. Physical Review Materials, 6(3), 033801.

Talks

SMC course: diffusion, LLMs, and SMC

February 28, 2025

Lecture at Linköping University, Liköping, Sweden

Learning on Graphs and Geometry Meetup Uppsala

February 12, 2025

Workshop at Uppsala University, Uppsala, Sweden

Teaching

Master’s thesis supervision

TDDE70 Deep Learning

TDDE01/732A99 Machine Learning

TDAB01 Probability and Statistics

Service and leadership

• Reviewer AISTATS 2024
• Reviewer AISTATS 2025