Every interesting system is evolving

We create models for learning dynamics and the dynamics of learning.
The goal is to make better forecasts and decisions.

News

  • Code release for Path-minimized Latent ODEs (jax)
  • Job opening: Postdoctoral researcher in Exoplanet Detection with Latent ODEs (Princeton University)
  • Blog post about the Latent ODE scheduler (code coming soon)

Papers

Wiemann, M.; Melchior, P.; Saydjari, A.
arXiv:2607.14466
Wiemann, M.; Smith, L.; Melchior, P.; Mishra-Sharma, S.; Wilson, A.; Izmailov, P.; Cuesta-Lázaro, C.
arXiv:2605.26087 (NeurIPS under review)
Krishnaraj, V.; Bayer, A.; Jespersen, C.; Melchior, P.
Journal of Cosmology and Astroparticle Physics, 2026, 026
Sampson, M.; Melchior, P.
arXiv:2509.23052
Yan, A.; Sampson, M.; Melchior, P.
NeurIPS 2025 TS4H workshop
Sampson, M.; Melchior, P.
Machine Learning: Science and Technology, 2025, 6, 025047