Inputs, Parameters, and Outputs

Every model computation has three roles.

The input is the data given to the model.

The parameters are the learned values stored by the model.

The output is the value computed by the model.

For a tiny model:

y_hat = wx + b

the roles are:

x       input
w, b    parameters
y_hat   output or prediction

The input is not learned. It is supplied.

The parameters are learned. They are changed by training.

The output is computed from the input and parameters.

One example

Suppose:

x = 3
w = 2
b = 1

Then:

y_hat = wx + b
      = 2 x 3 + 1
      = 7

The model predicted 7 for input 3.

DL-C02-T01-001Exercise: Compute a tiny prediction

Let y_hat = wx + b, with x = 4, w = 3, and b = -2. What is y_hat?

Your answer

Compute it first, then check your number.

HintSubstitute the values

Compute 3 x 4 + (-2).

SolutionWork it out

y_hat = 3 x 4 - 2 = 12 - 2 = 10.

DL-C02-T01-002Exercise: Name the learned values

In y_hat = wx + b, enter 1 if x is learned during training, or 0 if it is supplied as data.

Your answer

Compute it first, then check your number.

HintSeparate data from parameters

The model receives x. It stores and learns w and b.

SolutionWork it out

x is the input. It changes from example to example, but it is not learned by the model. The learned values are w and b.