DSC 261 Responsible Data Science Project
Collaborators:
- Adarsh Vemali
- Adhvaith Vijay
- Laurel Li
This project used the MNIST-1D dataset to delve into the Double Descent phenomenon in deep learning (specifically neural networks), and interpretatively analyzed how data points and model complexity affect model performance using SHAP, LIME and Saliency maps.
Baseline: This project was built using the PyTorch framework provided by Sam Greydanus of Windscape AI, which gave us the foundation for an implementation of 2 layer neural network model with varying number of neurons.