GitHub - erdc/ampl: Automated Machine-learning PipeLine
automlpipelineneural-networkhyperparameter-optimizationpython
Abstraction: ERDC automated ML pipeline for novice and expert users
Key points:
- AMPL (Automated Machine-learning PipeLine) was developed at the US Army Engineer Research and Development Center (ERDC) from work on ballistics kinetic energy reduction and rotorcraft airfoil drag coefficient prediction
- Workflow steps: Feature Importance, Hyperparameter optimization (via Optuna), Model training, Evaluation, and optional Ensemble
- Supports both neural networks (fully dense) and decision trees; results reported as percentage of predictions within 20/10/5/2.5/0.5% error thresholds
- Two interfaces: API (Python scripts via
Configuration+pipeline.run_all()) and CLI (python -m ampl config.yml) - Configuration driven by a YAML file specifying study name, target variable, feature list, and data path (CSV or SQLite)
- Designed for modularity: each pipeline step can be run independently or swapped out
Connections: Erdc · Automated Machine Learning · Hyperparameter Optimization · Neural Networks
Source: https://github.com/erdc/ampl