The authors introduce Mixture of GFlowNets (MoG), a community of K policies coupled through a jointly trained soft classifier.
Specialized GFlowNet mixtures find more modes without extra oracle calls
A classifier-coordinated population of samplers covered more high-reward regions than data-parallel training under the same environment-interaction budget.
Top University
Omar G. Younis · Joseph D. Viviano · Chirayu Anant Haryan · MD Vasimuddin · Sanchit Misra · Yoshua Bengio
Mila – Québec AI Institute · Intel · Université de Montréal
Research Digest··3 min read
Younis and colleagues replace a single Generative Flow Network with multiple independently trained GFlowNets encouraged to specialize in different regions of a compositional search space.
Why this paper
From Mila – Québec AI Institute and 2 others
In one line
A population of GFlowNets with a learned classifier divides state space to improve mode coverage beyond single-policy training.
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 (2 noted)
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§