Investigating a neural network response to input parameters using sensitivity analysis techniques.
Key Project Features
- Scalable
- Any number of features
- Higher order combinations
- Extensible
- New data
- New model
- New techniques
- Model Framework Independent
- Use Onnx
Core Techniques
- Boundary sensitivities
- One-Way sensitivity function
- Scenario decomposition
- Finite differences
Learning Outcome
- Understand essential mathematics for each technique.
- Build algorithm from existing formulation.
- Implement research paper.
- Analyze results.