Dmitrii Kochkov (@dkochkov1) profile picture

Discovering synergies between ML, computing and physics.

Cambridge, MA
Joined January 2021
Photo by Dmitrii Kochkov (@dkochkov1)Photo by Dmitrii Kochkov (@dkochkov1)Photo by Dmitrii Kochkov (@dkochkov1)Photo by Dmitrii Kochkov (@dkochkov1)
1/2 Our work "Learning ground states of quantum Hamiltonians with graph networks" is out! arxiv.org/abs/2110.06390 TLDR: we do Variational Monte Carlo on Heisenberg Hamiltonians using graph-nets as variational ansatz. w/ @spectralhippo Alvaro Sanchez @PeterWBattaglia @bkclark4
2
1
14
2/2 Key learnings: 1) Graph-nets learn amplitudes and phases that generalize to larger lattice sizes 2) Learning phases and amplitudes with separate networks gives better results 3) GNNs have easier time restoring symmetries than learning in a fully symmetric subspace
2
1/2 Excited to share "Machine learning accelerated computational fluid dynamics" arxiv.org/pdf/2102.01010.pdf We use ML inside a CFD simulator to advance the accuracy/speed Pareto frontier with/ Jamie A. Smith Ayya Alieva Qing Wang Michael P. Brenner @shoyer
12
114
17
497
2/2 We show that our models enable accurate simulations on coarse grids achieving 40-80x speedup on the same hardware while generalizing well to flows not seen during training!
4
4
1
60