Solutions

DL-C02-C-001

y_hat = 2 x 6 - 5
      = 12 - 5
      = 7

DL-C02-C-002

theta represents the model parameters. These are the learned values stored by the model.

DL-C02-C-003

The product has shape:

(12, 5) x (5, 4) -> (12, 4)

The bias has one value per output feature, so it broadcasts across the 12 examples. The final shape is (12, 4).

DL-C02-C-004

y is the target supplied by the dataset. y_hat is the model prediction.

DL-C02-C-005

prediction - target = 14 - 9 = 5

DL-C02-C-006

The architecture defines a function family. The parameters select one function from that family. Changing parameters changes the current function.