Neural networks from signals to decisions

Train the model. Watch the mechanism.

Change the architecture and training settings, then inspect how loss, accuracy, validation behaviour, and the learning rate move across epochs. Treat every run as an experiment with visible inputs and outputs.