analiza aiqbench
This research establishes a theoretical link between hierarchical variational inference and generative flow networks, demonstrating their equivalence in specific gradient contexts while highlighting GFlowNets' superior efficiency for off-policy training on discrete structures. The findings suggest that leveraging GFlowNet mechanisms could significantly improve diversity capture in multimodal distribution modeling without the high variance costs typical of importance sampling methods.
GFlowNets and variational inference
arXiv:2210.00580v4 Announce Type: replace Abstract: This paper builds bridges between two families of probabilistic algorithms: (hierarchical) variational inference (VI), which is typically used to model distributions over continuous spaces, and generative flow networks (GFlowNets), which have been used for distributions over discrete structures such as graphs. We demonstrate that, in certain cases, VI algorithms are equivalent to special cases of GFlowNets in the sense of equality of expected gradients of their learning objectives. We then point out the differences between the two families and show how these differences emerge experimentally. Notably, GFlowNets, which borrow ideas from reinforcement learning, are more amenable than VI to off-policy training without the cost of high gradient variance induced by importance sampling. We argue that this property of GFlowNets can provide advantages for capturing diversity in multimodal target distributions.
Abstract:This paper builds bridges between two families of probabilistic algorithms: (hierarchical) variational inference (VI), which is typically used to model distributions over continuous spaces, and generative flow networks (GFlowNets), which have been used for distributions over discrete structures such as graphs. We demonstrate that, in certain cases, VI algorithms are equivalent to special cases of GFlowNets in the sense of equality of expected gradients of their learning objectives. We then point out the differences between the two families and show how these differences emerge experimentally. Notably, GFlowNets, which borrow ideas from reinforcement learning, are more amenable than VI to off-policy training without the cost of high gradient variance induced by importance sampling. We argue that this property of GFlowNets can provide advantages for capturing diversity in multimodal target distributions.
Machine Learning (cs.LG); Machine Learning (stat.ML)
arXiv:2210.00580 [cs.LG]
(or arXiv:2210.00580v4 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2210.00580 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 224 cuvinte
Subiecte conexe
PhyloGFN: Phylogenetic inference with generative flow networks
arXiv · 1 sept. 2026
Variational Mixtures and Multi-Marginal Flow Matching: Advancing Statistical Inference with Biological Applications
arXiv · 30 sept. 2026
Sophisticated Policies from Epistemic Priors
arXiv · 23 iul. 2026
Seeing is Free, Speaking is Not: Uncovering the True Energy Bottleneck in Edge VLM Inference
arXiv · 19 aug. 2026