Bayesian Amortized Inference for Industrial Processes
Simulation-based Bayesian inference for fast uncertainty-aware parameters estimation in industrial process models.
This project explores Simulation Based Inference (SBI) a Bayesian amortized inference method for industrial process systems, using simulation-based inference to learn reusable probabilistic estimators from model-generated data. The goal is to support fast, uncertainty-aware inference for process monitoring, calibration, and decision support when direct likelihood evaluation is difficult or expensive.
The workflow combines prior sampling, process simulation, neural density estimation, and posterior inference so that repeated inference tasks can be handled efficiently after training.
