Optimizes and manages data preprocessing pipelines for BayesFlow simulation-based inference networks.
This skill provides expert guidance for architecting, debugging, and implementing BayesFlow adapters, which act as the bridge between raw simulator outputs and neural inference networks. It enforces essential domain-specific patterns, such as the three-slot schema mapping, the correct sequencing of transformations before standardization, and the use of analytical prior moments over empirical batch statistics to prevent distribution shifts. Whether you are implementing complex cross-key reparameterizations or declarative HPO-compatible specs, this skill ensures your data pipeline is robust and theoretically sound.
