Bridging Maximum Likelihood and Optimal Transport for Efficient Inference and Model Selection in Stochastic Block Models
arXiv:2605.28488v1 Announce Type: cross Abstract: We study inference in stochastic block models (SBMs) through the lens of optimal transport (OT). We first establish that maximum likelihood variational inference (MLVI) can be interpreted as a semi-relaxed Gromov-Wasserstein (srGW) projection with entropic regularization. While this formulation yields accurate clustering, the entropic regularization prevents transport plans to be sparse, hindering intrinsic model selection. Consequently, we investigate unregularized srGW estimators, and prove that they consistently recover both the SBM connecti
This research is emerging now as computational methods like optimal transport are being increasingly applied to statistical inference problems, especially in areas like machine learning and network analysis.
Improving inference and model selection in Stochastic Block Models (SBMs) contributes to more accurate and efficient analysis of complex network data, which is fundamental in many scientific and industrial applications.
The proposed unregularized semi-relaxed Gromov-Wasserstein (srGW) estimators offer a new, more intrinsically robust method for identifying community structures in networks compared to previous maximum likelihood variational inference approaches.
- · Machine learning researchers
- · Network scientists
- · Data analysts
More precise and reliable algorithms for community detection in networks will become available.
Applications relying on network analysis, such as social network understanding or bioinformatics, could see improvements in their underlying models.
These improved techniques might enable new insights in fields like epidemiology or financial risk modeling where understanding community structures is critical.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG