.

AbstractNN Documentation

AbstractNN is a Python library for formal verification of neural networks using abstract interpretation and affine arithmetic. It provides mathematically sound guarantees about network behavior under input perturbations.

Getting Started

• Overview
  • What is AbstractNN?
  • Key Features
  • Architecture
  • Core Components
  • Use Cases
  • Comparison with Other Methods
  • Next Steps
• Quick Start Guide
  • Basic Verification
  • Soundness Checking
  • Common Workflows
  • Configuration Tips
  • Next Steps
  • Common Issues
  • Need Help?

User Guide

• Command Line Interface
  • Overview
  • abstractnn-verify
  • abstractnn-eval
  • abstractnn-info
  • Common Workflows
  • Troubleshooting
  • See Also

API Reference

• API Reference
  • Core Modules
  • Module Overview
• Affine Engine
  • AffineExpression
  • AffineExpressionEngine
  • AffineExpression
  • AffineExpressionEngine
  • Examples
• Bound Propagator
  • BoundPropagator
  • BoundPropagator
  • Layer Handlers
  • Usage Examples

Indices and tables

• Index

• Module Index

• Search Page