BayesFlow - The Present and Future of Amortized Bayesian Inference
Details
Stefan Radev (LinkedIn profile) will discuss how to build and evaluate fully amortized neural Bayesian workflows within the BayesFlow framework. Specifically, he will focus on four key topics:
- posterior estimation
- model misspecification detection
- likelihood emulation
- model comparison
He will also showcase the new TensorFlow-based library for quickly specifying and validating amortized approximators designed for solving the above Bayesian tasks. Finally, he will discuss some major challenges and future vistas for amortized Bayesian inference.
Related topics
Machine Learning
Data Science
Python
Open Source
Bayesian Statistics