The framework begins with an irreducible, reversible Markov kernel whose transitions define which states can exchange probability.
Discrete Wasserstein transport enables one-step generation on finite state spaces
The authors construct a KL-decreasing Markov jump flow and train a neural generator to reproduce its transport in one forward pass.
Top University
Alessandro Micheli · Andrea Zerio · Samir Bhatt
Imperial College London · Aalborg University · Centre for Frontier AI Research (CFAR) · Institute of Advanced Intelligence and Computing (IAIC) · A*STAR
Research Digest··2 min read
Micheli, Zerio and Bhatt propose Discrete Drifting, a framework that adapts one-step generative modeling to finite state spaces.
Why this paper
From A*STAR and 5 others
In one line
Discrete Drifting uses discrete Wasserstein geometry to enable one-step generative modeling on finite state spaces via Markov jump processes.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§