Reproduce “Kolmogorov-Chapman Maps: Trajectory-Free Learning for Simulating Molecular Dynamics” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Majdi Hassan, Marta Skreta, Dominique Beaini, Kirill Neklyudov

KC-Maps learns molecular-dynamics transitions across larger time intervals from unpaired equilibrium configurations and known force-field drift and diffusion. It trains transitions to compose consistently and match short-time dynamics, rather than relying on paired trajectory states. On AL6, after one GPU-hour on a single L40S, its TICA-W₂ was 0.60 ± 0.03, versus 1.01 ± 0.02 for MD; this result is system- and metric-specific.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.