Articial Neural Networks & Deep Learning

1.

Multiple Choice

30 sec

1 pt

A neural network trained with labelled data is an example of . . .

supervised learning

unsupervised learning

2.

Multiple Choice

45 sec

1 pt

In the neural network shown,

each neuron is connected to a random selection of other neurons in the network

every neuron is connected to every other neuron in the network

every neuron is conected to every neuron in the next layer

each neuron is connected to a random selection of neurons in the next layer

3.

Multiple Choice

1 min

1 pt

The activation of each neuron in a layer is determined by the __________ of the activations of all the neurons in the previous layer.

weighted sum

sum

level

4.

Multiple Choice

1 min

1 pt

Which of these equations means:

"Output O is equal to zero if input I is less than zero; or equal to input I if I is greater than or equal to zero"

?

5.

Multiple Choice

1 min

1 pt

Which of these equations means:

"Output O is always greater than or equal to zero, and less than or equal to 1, whatever the input"

?

6.

Multiple Choice

1 min

1 pt

Which of these corresponds to the output of a sigmoid activation function used for artificial neural networks?

7.

Multiple Choice

1 min

1 pt

Which of these corresponds to the output of the ReLu activation function used for Deep Learning networks?

8.

Multiple Choice

2 mins

1 pt

Consider this diagram of a perceptron with an input layer, hidden layer and output layer.

The activations of neurons i1, i2, and i3 of the Input layer are, respectively:

0.3, 0.5, -0.5

0.4, 0.2, 0.2

9.

Fill in the Blank

5 mins

1 pt

Activation of:

i1 = 0.4, i2 = 0.2, i3 = 0.2

Connection weight to h2 from:

i1 = 0.3, i2 = 0.5, i3 = -0.5

What is the activation of h2? _

10.

Multiple Choice

5 mins

1 pt

11.

Multiple Choice

5 mins

1 pt

12.

Multiple Choice

45 sec

1 pt

Remember vectors and scalars? A scalar is a single number; a vector is an ordered collection of numbers.

Pick the scalar:

13.

Fill in the Blank

30 sec

1 pt

14.

Fill in the Blank

30 sec

1 pt

15.

Fill in the Blank

30 sec

1 pt

16.

Fill in the Blank

30 sec

1 pt

17.

Fill in the Blank

30 sec

1 pt

18.

Multiple Choice

30 sec

1 pt

(btw, there are ~530m neurons in a dog's brain - I made a mistake in the moodle quiz.)

Which has the most neurons?

dog

frog

human

19.

Multiple Choice

2 mins

1 pt

If a = actual activation and t = target activation,

the cost c for a single output neuron is given by:

20.

Multiple Choice

2 mins

1 pt

The sum of the cost c for all the output neurons

The mean of the cost c for all the output neurons

21.

Multiple Choice

2 mins

1 pt

Backpropagation of a neural network is undertaken on the basis of:

the cost C of a single run of the network with a single input example

the mean of the costs C from many runs of the network with different inputs examples

22.

Multiple Choice

1 min

1 pt

The cost function for backpropagation of a neural network provides ...

a vector of values as a measure of error

a single scalar as a measure of error

23.

Multiple Choice

30 sec

1 pt

Early neural networks had ...

a few layers

many layers

24.

Multiple Choice

30 sec

1 pt

Deep learning networks may comprise . . .

up to 10 layers

up to 50 layers

100 layers of more

25.

Multiple Choice

1 min

1 pt

The size of networks is measured by how many parameters it has - that is, the number of connection weights which must be adjusted in learning. GPUs (graphical processing units) are often used in this context, because they

... have parallel processing, which makes the process faster

... use less electricity than CPUs

Explore this activity with a free account