Hello everyone,
I am doing the machine learning course and was wondering how do ou decide number of filters, dropout and flatten when we build our model?
Basically, you experiment and see what works and what doesn’t. Or you can read about similar classification networks and see what worked for them as a starting point. Or you can read about standard classification networks to see how they are structured and simplify it to run in this project.
You’re constrained on the data you have for testing and training, so really your only experimental variable is the layer selection.