Deep Learning — Quiz | Data Science using Python

Question 1

What does a single neuron compute?

A A median

B A weighted sum of inputs plus bias, then an activation

C A correlation matrix

D A random number

Question 2

Why are nonlinear activation functions essential?

A They speed up training

B Without them, stacked layers collapse into a single linear model

C They remove outliers

D They are required for plotting

Question 3

What is the ReLU activation?

A 1 / (1 + e^-x)

B max(0, x)

C x squared

D the mean of x

Question 4

What does the forward pass produce?

A Updated weights

B A prediction and the loss

C The gradients

D The dataset

Question 5

What does backpropagation compute?

A The final accuracy

B The gradient of the loss with respect to each weight

C The number of epochs

D The learning rate

Question 6

In Keras, which trio defines the training setup?

A read, clean, save

B Sequential layers, compile, fit

C fork, merge, push

D scale, encode, split

Question 7

What does Dropout do?

A Deletes rows of data

B Randomly mutes neurons during training to reduce overfitting

C Increases the learning rate

D Adds more layers

Question 8

What does EarlyStopping do?

A Stops on the first epoch always

B Halts training when validation loss stops improving and restores the best weights

C Removes the test set

D Doubles the batch size

Question 9

Why do CNNs outperform dense networks on images?

A They use no weights

B They exploit spatial structure with sliding filters

C They are always smaller

D They need no training

Question 10

What does a pooling layer do in a CNN?

A Adds colour

B Reduces spatial size while keeping the strongest features

C Increases resolution

D Labels the image

Question 11

What is transfer learning?

A Copying files between computers

B Reusing a model pretrained on large data and adapting it to your task

C Training from scratch

D Moving data to the cloud

Question 12

In basic transfer learning, what do you do with the pretrained backbone first?

A Delete it

B Freeze it and train only a small new head

C Retrain all of it from scratch

D Ignore its weights