ECNet documentation

ECNet is a Python package for building multilayer-perceptron models that predict fuel properties from molecular structure using quantitative structure–property relationship (QSPR) descriptors. It ships bundled property datasets, PaDEL and alvaDesc descriptor backends, hyperparameter-tuning helpers, and analytical blend-property equations.

This site documents the current v4 API. The 2017 JOSS article and paper/paper.md describe a prior project/build/node ensemble architecture; that workflow is historical and is not the public surface documented here.

Contents

• Installation
  • Install from PyPI
  • Upgrade
  • Install from source
  • Development extras
• Quickstart
  • Train a small model
  • Save and reload
  • Next steps
• Units and numeric conventions
  • Blend properties
  • Model targets
• Bundled property datasets
  • Common attributes
  • Property cards
  • Repository research archive (databases/)
• API stability policy
  • Frozen public surface
  • Advanced import: PCADataset
  • Versioning policy
  • Executable contract
• API reference
  • ecnet
  • ecnet.model
  • ecnet.datasets
  • ecnet.tasks
  • ecnet.blends
  • ecnet.callbacks

Indices and tables

• Index

• Module Index

• Search Page