Deep Learning

Question 1

What functions can make up a neural network

Answer 1

Simple functions like convolutions, filtering, resizing etc.

Question 2

Convolutional neural network

Answer 2

A neural network that learns the values of the kernels that it uses to convolve with the image

Question 3

Max Pooling

Answer 3

A neuron that returns the highest value of its input

Question 4

Deconvolution

Answer 4

A type of convolution that makes the image larger

Question 5

Skip connections

Answer 5

Connections between non-consecutive layers of a neural network - useful for retaining information about images without losing it

Question 6

Supervised learning

Answer 6

Learning that uses labelled datasets

Question 7

Loss function

Answer 7

The function that calculates how incorrect the output of a network was

Question 8

Backpropagation

Answer 8

We use the chain rule to take the derivative of each function in the neural network, this helps approximate the loss function and train the network accordingly

Question 9

Ground truth

Answer 9

The guaranteed correct output of an input for a neural network

Question 10

Adversarial neural network

Answer 10

Two networks work to generate an output
The generator network generates output
The pre-trained discriminator network tries to distinguish between ground truth data and data generated by the generator - once it can’t tell the difference, the generator is generating good data