PERCEPTRONIUM

Perceptronium is a lightweight toolkit for building Behler–Parrinello symmetry-function descriptors and training neural-network interatomic potentials. It combines a fast C++ descriptor-generation core with Python utilities for data conversion, packaging, and model training, making it easy to move from atomistic trajectories and energies to trainable ML-ready inputs.

Documentation:

• Getting started
• Usage
  • Assumptions
  • Step 1: Prepare geometry and energy files
  • Step 2: Convert to a package file
  • Step 3: Create the request file
  • Step 4: Generate descriptors
  • Step 5: Fit a neural network
  • Summary workflow
• Tutorials
  • Tutorial: Creating a neural network for methane
• Symmetry Functions
  • One-body functions
  • Two-body functions

Indices and tables

• Index

• Module Index

• Search Page