Deep learning basics

Question 1 of 8

Backpropagation computes:

Question 2 of 8

Without a non-linear activation, a deep network is equivalent to:

Question 3 of 8

The vanishing gradient problem is worst with:

Question 4 of 8

Dropout works by:

Question 5 of 8

A learning rate that is too high causes:

Question 6 of 8

Batch normalization primarily:

Question 7 of 8

A convolutional layer is well suited to images because it:

Question 8 of 8

If training loss falls but validation loss rises, you should first: