Free Practice Quiz Question List

5 Neural Networks Online Quiz Questions

Use this free practice quiz with 20 questions to review 5 Neural Networks, test your knowledge, and prepare for your next test or exam.

20 questions

Choose one

1 point

A network receives measurements as features and produces a numerical estimate. Which description correctly identifies the roles of its input and output layers?

A
The input layer receives the measurements, and the output layer produces the prediction.
B
The input layer produces the prediction, and the output layer receives the measurements.
C
Hidden layers receive the correct targets, and the output layer adjusts the weights.
D
The output layer transforms information between the input and hidden layers.

True or false

1 point

Artificial neurons are literal copies of biological brain cells.

A
True
B
False

Fill in the blank

1 point

A neuron multiplies each input by a learned and adds a learned before applying its activation function.

Written response

1 point

A training run has completed one full pass through all of its training examples. What is this full pass called?

Fill in the blank

1 point

In an image-recognition model, early layers may respond to , while later layers combine them into .

Choose one

1 point

A model needs an output that can represent a probability for a binary prediction. Why might a sigmoid activation be suitable for that output?

A
To ensure every hidden layer is linear.
B
To represent a numerical estimate without restricting its range.
C
To represent a probability for a binary prediction.
D
To guarantee that the model generalizes to unseen data.

Choose all

1 point

Select all statements that correctly describe the output ranges of these activation functions.

A
Sigmoid maps values between zero and one.
B
Tanh maps values between negative one and one.
C
Tanh maps values only between zero and one.
D
Sigmoid always returns either exactly zero or exactly one.

True or false

1 point

Stacking layers that perform only linear calculations is enough, by itself, to let a network represent nonlinear patterns.

A
True
B
False

True or false

1 point

Evaluating a trained network on separate, unseen data helps assess whether its learned patterns generalize beyond the training examples.

A
True
B
False

Choose all

1 point

Select all statements supported about deep neural networks.

A
A deep neural network has multiple hidden layers.
B
Its layers can build successively transformed representations of the input.
C
More layers and neurons can increase the number of parameters and training demands.
D
Every layer is guaranteed to have a neat, human-readable meaning.

Choose one

1 point

During supervised training, what is the primary role of the loss function?

A
To initialize weights and biases to small values.
B
To compare predictions with target values and measure error.
C
To pass values from one layer to the next.
D
To guarantee that the learned patterns generalize.

Open ended

1 point

Explain how backpropagation and an optimizer work together during training to improve a neural network's predictions.

Choose one

1 point

A neural network receives measurements as input and predicts a class. Which layer produces the prediction?

A
The input layer receives the original features.
B
The output layer produces the network’s prediction.
C
A hidden layer adjusts the weights during training.
D
A loss function measures the prediction error.

Choose one

1 point

Why do layered neural networks use activation functions between linear calculations?

A
They make each layer require fewer parameters.
B
They let the network represent nonlinear patterns.
C
They ensure every prediction matches its target.
D
They remove the need to train the weights.

Choose one

1 point

A model needs an output that can represent a probability for a binary prediction. Which activation function can be suitable for that output?

A
ReLU, because it maps every input to a value between 0 and 1.
B
Tanh, because it maps every input to a probability between 0 and 1.
C
Sigmoid, because it maps values between 0 and 1.
D
Any activation function, because all activations have the same range.

Choose one

1 point

During supervised training, what does a loss function measure?

A
It measures how predictions compare with target values.
B
It carries values from one layer to the next.
C
It initializes the network's weights and biases.
D
It determines how many layers the network has.

Choose one

1 point

After gradients have been calculated, what does an optimizer commonly do during training?

A
It compares predictions with targets to measure error.
B
It passes input values through the network to produce predictions.
C
It calculates the initial values of all targets.
D
It uses gradients to update parameters in a direction that reduces loss.

Written response

1 point

A neuron computes $z = w_1x_1 + w_2x_2 + b$ . If $x_1 = 2$ , $x_2 = -1$ , $w_1 = 0.5$ , $w_2 = 3$ , and $b = -2$ , what is the value of $z$ ?

Written response

1 point

What is the term for one full pass through the training examples?

Written response

1 point

In a common network design, every neuron in one layer connects to every neuron in the next. If one layer has 4 neurons and the next has 3, how many connections run between these layers?