syz
@syz.bsky.social
๐ค 14
๐ฅ 28
๐ 0
reposted by
syz
Marco Cuturi
6 days ago
We also introduce two coupling approaches advocated this summer to improve FM training: using either very large sharp Sinkhorn couplings (
arxiv.org/abs/2506.05526
) or, even better, semidiscrete couplings (
arxiv.org/abs/2509.25519
), as proposed with Alireza Mousavi-Hosseini and
@syz.bsky.social
loading . . .
On Fitting Flow Models with Large Sinkhorn Couplings
Flow models transform data gradually from one modality (e.g. noise) onto another (e.g. images). Such models are parameterized by a time-dependent velocity field, trained to fit segments connecting pai...
https://arxiv.org/abs/2506.05526
0
1
1
reposted by
syz
Marco Cuturi
6 days ago
We have been working with Michal Klein on pushing a module to train *flow matching* models using JAX. This is shipped as part of our new release of the OTT-JAX toolbox (
github.com/ott-jax/ott
) The tutorial to do so is here:
ott-jax.readthedocs.io/tutorials/ne...
loading . . .
1
11
6
reposted by
syz
David Picard
about 1 month ago
Wow! Finally OT done on the entire training set to train a diffusion model!
add a skeleton here at some point
0
12
3
reposted by
syz
Marco Cuturi
about 1 month ago
Our two phenomenal interns, Alireza Mousavi-Hosseini and Stephen Zhang
@syz.bsky.social
have been cooking some really cool work with Michal Klein and me over the summer. Relying on optimal transport couplings (to pick noise and data pairs) should, in principle, be helpful to guide flow matching ๐งต
2
30
8
reposted by
syz
arXiv cs.LG Machine Learning
about 1 month ago
Alireza Mousavi-Hosseini, Stephen Y. Zhang, Michal Klein, Marco Cuturi: Flow Matching with Semidiscrete Couplings
https://arxiv.org/abs/2509.25519
https://arxiv.org/pdf/2509.25519
https://arxiv.org/html/2509.25519
0
2
2
you reached the end!!
feeds!
log in