Resources

We share lightweight, hands-on resources (Jupyter/Pluto notebooks, tutorials, and code) so you can explore AI verification and validation topics directly in your browser via Google Colab.

Notebooks

Interactive Notebooks

ML Robustness for Demand Prediction

Demonstrates that some ML models are more robust than others — small input changes lead to small output changes. Compares classical models (via PyCaret) with FFNN, RNN, and LSTM (via PyTorch).

Robustness PyCaret PyTorch Time Series

Adversarial Attack on Traffic Light Classification

Demonstrates how a small, imperceptible perturbation can fool a neural network into misclassifying a traffic light as a tree. Uses Foolbox with a pretrained ResNet-18 model on ImageNet.

Adversarial Attacks Foolbox ResNet AI Safety

NASA V&V: Robot Trajectory Verification

Operationalizes NASA's Product Verification Process (Chapter 5.3) in a Jupyter notebook. Builds a neural network for robot trajectory prediction, then applies the full five-activity verification harness with structured requirements, traceability, and reporting.

NASA Verification Systems Engineering PyTorch

All notebooks are open source and available on GitHub .