Neural networks explained simply quiz

Question 1

What is a neural network in simple words?

Accepted Answer

A computer system of connected units that learns patterns from examples

Answer

A type of internet cable that connects brains

Answer

A spreadsheet that stores numbers

Answer

A robot with a real human brain inside

Question 2

What is an artificial neuron?

Accepted Answer

A unit that combines weighted inputs and produces an output

Answer

A real brain cell placed inside a computer

Answer

A picture stored in memory

Answer

A button you press to start the computer

Question 3

What is a 'weight' in a neural network?

Accepted Answer

A number showing how important an input is

Answer

The physical mass of the computer

Answer

The number of neurons in a layer

Answer

The speed of the internet connection

Question 4

What does a 'bias' do in a neuron?

Accepted Answer

An extra number that shifts a neuron's output

Answer

A way the network unfairly favors some people

Answer

The error made by the network

Answer

The number of layers used

Question 5

What are the three main types of layers in a neural network?

Accepted Answer

Input, hidden, and output layers

Answer

Top, middle, and bottom layers

Answer

Red, green, and blue layers

Answer

Start, pause, and stop layers

Question 6

Why are they called 'hidden' layers?

Accepted Answer

They sit between input and output, so we do not directly see their values

Answer

They are encrypted for security

Answer

They are stored on a hidden hard drive

Answer

They only work at night

Question 7

What is an activation function?

Accepted Answer

A function that decides a neuron's output and adds non-linearity

Answer

A button that turns the network on

Answer

A function that deletes unused neurons

Answer

A tool that draws charts

Question 8

What does the ReLU activation function do?

Accepted Answer

Keeps positive numbers and turns negatives into zero

Answer

Doubles every number

Answer

Turns positives into zero and keeps negatives

Answer

Sorts numbers from low to high

Question 9

What is the sigmoid function and what range does its output have?

Accepted Answer

It squashes inputs to a value between 0 and 1

Answer

It outputs only whole numbers

Answer

It outputs values between -100 and 100

Answer

It always outputs exactly 0.5

Question 10

What does it mean to 'train' a neural network?

Accepted Answer

Adjusting weights using examples so predictions improve

Answer

Plugging the computer into power

Answer

Drawing the network on paper

Answer

Cleaning the computer screen

Question 11

What is a 'loss function' (or error) in training?

Accepted Answer

A measure of how wrong the network's predictions are

Answer

The amount of electricity used

Answer

The number of neurons lost during training

Answer

The time taken to start the computer

Question 12

What is backpropagation, in simple terms?

Accepted Answer

Sending the error backward to update each weight

Answer

Copying data to a backup drive

Answer

Turning the network off and on again

Answer

Drawing the network from back to front

Question 13

What is gradient descent used for?

Accepted Answer

To adjust weights step by step to reduce the loss

Answer

To increase the number of layers

Answer

To draw graphs of data

Answer

To delete bad neurons

Question 14

What is a 'learning rate'?

Accepted Answer

How big each weight-update step is

Answer

How many students learn at once

Answer

The speed of the internet

Answer

The number of correct answers

Question 15

What is an 'epoch' in training?

Accepted Answer

One full pass through the whole training dataset

Answer

A single neuron in the network

Answer

The final answer of the network

Answer

A type of activation function

Question 16

What is the difference between training data and test data?

Accepted Answer

Training data is learned from; test data checks new performance

Answer

They are exactly the same thing

Answer

Test data is larger and used for learning

Answer

Training data is only used after testing

Question 17

What does 'overfitting' mean?

Accepted Answer

Memorizing training data so new data is handled poorly

Answer

Learning too few examples

Answer

Using too small a computer

Answer

Training for too short a time

Question 18

What is 'underfitting'?

Accepted Answer

The network is too simple and learns the pattern poorly

Answer

The network memorizes everything perfectly

Answer

The network has too many layers

Answer

The data is too clean

Question 19

How is a neural network inspired by the human brain?

Accepted Answer

It uses connected artificial neurons, like brain cells, but much simpler

Answer

It contains real living brain cells

Answer

It is exactly as powerful as a human brain

Answer

It has nothing to do with the brain

Question 20

What is a 'deep' neural network?

Accepted Answer

A network with many hidden layers

Answer

A network buried deep underground

Answer

A network that uses deep blue colors

Answer

A very old type of network

Question 21

What kind of tasks are neural networks good at?

Accepted Answer

Recognizing patterns like images, speech, and language

Answer

Only doing simple addition

Answer

Storing files like a folder

Answer

Charging a phone battery

Question 22

Why do neural networks need a lot of data?

Accepted Answer

More examples help them learn accurate, general patterns

Answer

Data makes the screen brighter

Answer

They eat data for power

Answer

Data is only needed for storage

Question 23

What is a Convolutional Neural Network (CNN) mainly used for?

Accepted Answer

Working with images, like recognizing objects in photos

Answer

Only adding numbers

Answer

Storing music files

Answer

Controlling the mouse pointer

Question 24

What is a Recurrent Neural Network (RNN) good at?

Accepted Answer

Handling sequences like text and time series using memory

Answer

Only working with single still images

Answer

Cleaning datasets

Answer

Making the computer faster

Question 25

In image recognition, how does a neural network 'see' a picture?

Accepted Answer

As a grid of numbers representing pixels

Answer

As a printed photograph

Answer

As written words describing it

Answer

As a sound recording

Question 26

What is 'supervised learning' for a neural network?

Accepted Answer

Learning from examples that come with correct labels

Answer

Learning with no data at all

Answer

Learning only by playing games

Answer

Learning by copying another computer

Question 27

What is the difference between supervised and unsupervised learning?

Accepted Answer

Supervised uses labeled data; unsupervised finds patterns without labels

Answer

Both always use labeled data

Answer

Unsupervised is faster internet

Answer

Supervised needs no data

Question 28

What does a single neuron compute, written as a formula?

Accepted Answer

activation(weights times inputs, plus bias)

Answer

inputs minus weights only

Answer

just the largest input

Answer

the average of all weights

Question 29

Why is non-linearity important in neural networks?

Accepted Answer

It lets the network learn complex, non-straight-line patterns

Answer

It makes the computer use less power

Answer

It deletes extra neurons

Answer

It speeds up the internet

Question 30

What is a 'parameter' in a neural network?

Accepted Answer

The weights and biases the network learns

Answer

The colors on the screen

Answer

The keyboard keys used

Answer

The brand of the computer

Test yourself