SIMPLE-NN

Official Resources

• Homepage: https://github.com/MDIL-SNU/SIMPLE-NN
• Documentation: https://github.com/MDIL-SNU/SIMPLE-NN
• Source Repository: https://github.com/MDIL-SNU/SIMPLE-NN
• License: Apache License 2.0

Overview

SIMPLE-NN (SImple Machine Learning Potential Energy with Neural Networks) is a package for constructing neural network potentials using Behler-Parrinello symmetry functions. It is designed to be easy to use and integrates with LAMMPS for MD simulations.

Scientific domain: Machine learning potentials
Target user community: MD users

Capabilities (CRITICAL)

• Descriptors: Atom-centered symmetry functions (ACSF).
• Model: Feed-forward neural networks.
• LAMMPS: Interface provided.
• Google Colab: Tutorials available.

Sources: SIMPLE-NN GitHub, Comp. Phys. Comm. 220, 158 (2017)

Inputs & Outputs

• Input formats: VASP OUTCAR, XYZ
• Output data types: Potential files

Interfaces & Ecosystem

• TensorFlow: Backend.
• LAMMPS: MD engine.

Workflow and Usage

1. Prepare data.
2. Train model.
3. Run LAMMPS MD.

Performance Characteristics

• Standard NN performance.
• Easy entry point for beginners.

Application Areas

• General MD
• Phase transitions

Community and Support

• Developed by Han Group (Seoul National University)

Verification & Sources

Primary sources:

1. GitHub: https://github.com/MDIL-SNU/SIMPLE-NN
2. Publication: K. Lee et al., Comp. Phys. Comm. 220, 158 (2017)

Confidence: VERIFIED

Verification status: ✅ VERIFIED

• Website: ACTIVE
• Documentation: AVAILABLE
• Source: OPEN (GitHub)
• Development: ACTIVE
• Applications: NN potentials