A collection of Jupyter notebooks to help newcomers understand FastAI, hyperparameters, and more.
Dropout Rates - Blog Post
- Testing Dropout Rates (small images) - fully annotated results for comparing dropout rates from 0 to 0.9 (0 - 90% dropout).
- Testing Dropout Rates (large images) - no additional notes but a demonstration of how loss is a more useful chart on larger images (and how it takes notably longer).
- Testing Dropout Rates (specific layers) - no additional notes but a notebook exploring dropout in different layers.
Created by Mario Lurig https://mariolurig.com/