ReLU

Sigmoid

Tanh

Softmax

D

A

C

B

(32,)

(10, 32)

(1,)

(10,)

A binary classifier using RNN

An RNN that outputs a sequence

A CNN for image classification

A regression model using CNN

(32,32,32)

(3,3,,32)

(1,1,32)

(28,28,32)

(34, 34, 64)

(32, 32, 64)

(3, 3, 3)

(30, 30, 64)

Apply dropout

Pool feature maps

Extract features from input

Perform classification

4

1

2

Doubles the output layer size

Randomly ignores 50% of the neurons during training

 Halves the learning rate

Removes 50% of inputs permanently

It adds parameters to the model

It changes the output shape

It has trainable parameters