Two-way Guidance: Simultaneous Learning of Summary Statistics for Generative Bayes

Key points

• Bayesian simulation-based inference (SBI) can struggle to recover increasingly concentrated posteriors as observation size grows.1
• Our approach instead imposes structure on how the posterior learner depends on a learned summary (e.g., affinity), and learns the summary simultaneously with the posterior.1
• We identify an important mechanism named two-way guidance: under a structural restriction on the summary, the posterior sufficiency of the learned summary simplifies posterior learning, whereas a suitable posterior objective guides the learned summary towards posterior sufficiency.1
• We establish theoretical guarantees on both posterior estimation error and the learned summary's departure from sufficiency, thus identifying conditions on the simulation budget for recovering increasingly concentrated posteriors as observation size grows.1

Sources (1)

• [1]Two-way Guidance: Simultaneous Learning of Summary Statistics for Generative Bayes
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 01:16 AM

  We consider learning sufficient summary statistics jointly with the posterior as a general principle for scalable inference in various SBI methods.

  Bayesian simulation-based inference (SBI) can struggle to recover increasingly concentrated posteriors as observation size grows.

Extractive summary: sentences quoted from the sources.