What is a BayesFlow adapter?

A BayesFlow adapter is a data preprocessing pipeline that transforms raw simulator outputs into the specific data schema (inference, summary, and conditions) required by BayesFlow networks.

What is the correct order for an adapter pipeline?

The recommended order is: 1. Type conversion, 2. Cross-key transforms, 3. Per-key transforms, 4. Standardization (after transforms), 5. Schema concatenation/renaming, and 6. Key filtering.

How do I handle multi-variable reparameterization?

You can subclass the Transform class for cross-key operations. These should be added early in the forward pipeline so they see original-scale parameters during the inverse path.

Why should I standardize using prior moments instead of empirical statistics?

Using analytical prior moments ensures deterministic and constant standardization. This prevents the distribution shifts that often occur when using empirical batch statistics, especially with variable-size data.

BayesFlow Adapter Orchestrator

Name: BayesFlow Adapter Orchestrator
Author: matthiaskloft

bymatthiaskloft

0•

数据科学与机器学习

Streamlines the creation and debugging of data preprocessing pipelines for BayesFlow simulation-based inference.

This skill provides specialized assistance for building BayesFlow adapters, which are essential for mapping simulator outputs into structured schemas for summary and inference networks. It helps developers implement best practices, such as prioritizing transformations before standardization and using analytical prior moments instead of empirical statistics to avoid distribution shifts. Whether you are defining inference variables, handling complex reparameterizations via cross-key transforms, or configuring HPO-compatible AdapterSpecs, this skill ensures your data pipeline is robust, reproducible, and correctly ordered.

主要功能

01Automated pipeline construction with enforced operation ordering

02Mapping logic for inference, summary, and conditioning variables

03Declarative AdapterSpec patterns for hyperparameter optimization

04Support for custom elementwise and cross-key reparameterization transforms

05Standardization using analytical prior moments to prevent training shifts

060 GitHub stars

使用场景

01Debugging distribution shifts caused by incorrect standardization logic

02Reparameterizing simulator outputs for improved neural network convergence

03Designing preprocessing pipelines for simulation-based inference models

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add matthiaskloft/claude-skills bayesflow-adapter

For use in Claude.ai and ChatGPT

主要功能

01Automated pipeline construction with enforced operation ordering

02Mapping logic for inference, summary, and conditioning variables

03Declarative AdapterSpec patterns for hyperparameter optimization

04Support for custom elementwise and cross-key reparameterization transforms

05Standardization using analytical prior moments to prevent training shifts

060 GitHub stars

使用场景

01Debugging distribution shifts caused by incorrect standardization logic

02Reparameterizing simulator outputs for improved neural network convergence

03Designing preprocessing pipelines for simulation-based inference models

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add matthiaskloft/claude-skills bayesflow-adapter

For use in Claude.ai and ChatGPT