Phase 5: Neural Networks - Pre-Quiz

Time: 15 minutes
Questions: 10
Passing Score: 70%
Purpose: Assess your baseline knowledge before learning about neural networks

Question 1 (Easy)

What is a neural network inspired by?

A) Computer circuits
B) The human brain ✓
C) Mathematical functions
D) Decision trees

Explanation

Answer: B) The human brain

Neural networks are loosely inspired by biological neurons in the brain. They consist of interconnected nodes (neurons) that process information, similar to how biological neurons fire signals.

Reference: Phase 5 - Introduction to Neural Networks

Question 2 (Easy)

What is the basic unit of a neural network called?

A) Cell
B) Node
C) Neuron ✓
D) Layer

Explanation

Answer: C) Neuron

While “node” is also commonly used, the formal term is “neuron” or “artificial neuron,” reflecting the biological inspiration.

Reference: Phase 5 - Neural Network Basics

Question 3 (Medium)

In a neural network, what does a “weight” represent?

A) The size of the network
B) The strength of a connection between neurons ✓
C) The number of neurons
D) The learning rate

Explanation

Answer: B) The strength of a connection between neurons

Weights are parameters that determine how much influence one neuron has on another. During training, these weights are adjusted to improve the network’s performance.

Reference: Phase 5 - Weights and Biases

Question 4 (Medium)

What is the purpose of an activation function?

A) To speed up training
B) To introduce non-linearity ✓
C) To reduce overfitting
D) To normalize inputs

Explanation

Answer: B) To introduce non-linearity

Activation functions introduce non-linearity into the network, allowing it to learn complex patterns. Without activation functions, a neural network would just be a linear regression model regardless of how many layers it has.

Reference: Phase 6 - Activation Functions

Question 5 (Easy)

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

A) Input, Output, Middle
B) Input, Hidden, Output ✓
C) Start, Process, End
D) Data, Compute, Result

Explanation

Answer: B) Input, Hidden, Output

• Input layer: Receives the initial data

• Hidden layer(s): Process the data (can be multiple)

• Output layer: Produces the final prediction

Reference: Phase 5 - Network Architecture

Question 6 (Hard)

What is backpropagation used for?

A) Forward pass computation
B) Calculating gradients for weight updates ✓
C) Initializing weights
D) Preventing overfitting

Explanation

Answer: B) Calculating gradients for weight updates

Backpropagation is an algorithm that calculates the gradient of the loss function with respect to each weight using the chain rule of calculus. These gradients are then used to update the weights via gradient descent.

Reference: Phase 5 - Backpropagation

Question 7 (Medium)

What does the “learning rate” control?

A) How fast the network learns
B) The size of weight updates ✓
C) The number of epochs
D) The batch size

Explanation

Answer: B) The size of weight updates

The learning rate (α) determines how much the weights are adjusted during each training step. Too large: training becomes unstable. Too small: training is very slow.

Formula: new_weight = old_weight - learning_rate * gradient

Reference: Phase 5 - Training Hyperparameters

Question 8 (Hard)

import numpy as np
weights = np.array([0.5, -0.3, 0.8])
inputs = np.array([1, 2, 3])
bias = 0.1
output = np.dot(weights, inputs) + bias

What is the value of output?

A) 1.4
B) 2.0
C) 2.4 ✓
D) 3.0

Explanation

Answer: C) 2.4

Calculation:

output = (0.5 × 1) + (-0.3 × 2) + (0.8 × 3) + 0.1
       = 0.5 + (-0.6) + 2.4 + 0.1
       = 2.4

This is a simple neuron computation: weighted sum of inputs plus bias.

Reference: Phase 5 - Forward Pass

Question 9 (Medium)

What is the main difference between a shallow and a deep neural network?

A) Input size
B) Number of hidden layers ✓
C) Activation function used
D) Learning rate

Explanation

Answer: B) Number of hidden layers

• Shallow network: 1-2 hidden layers

• Deep network: 3+ hidden layers (hence “deep learning”)

Deep networks can learn more complex, hierarchical representations of data.

Reference: Phase 5 - Deep Learning Basics

Question 10 (Hard)

Which statement about the vanishing gradient problem is TRUE?

A) It only affects the output layer
B) It makes gradients in early layers very small ✓
C) It speeds up training
D) It only occurs with ReLU activation

Explanation

Answer: B) It makes gradients in early layers very small

In deep networks with certain activation functions (e.g., sigmoid, tanh), gradients can become extremely small as they propagate backward through layers. This makes it difficult to update weights in early layers.

Solution: Use ReLU activation or batch normalization

Reference: Phase 5 - Training Challenges

Scoring Guide

0-3 correct (0-30%): Don’t worry! Neural networks are advanced. Start with the basics and take your time.

4-5 correct (40-50%): You have some foundation. Review mathematical concepts (linear algebra, calculus) before diving into neural networks.

6-7 correct (60-70%): Good baseline! You’re ready to learn neural networks with some effort.

8-9 correct (80-90%): Excellent! You have strong fundamentals. This phase will solidify your knowledge.

10 correct (100%): Outstanding! You may already know this content, but you’ll still learn implementation details and best practices.

Next Steps

After taking this pre-quiz:

1. Score < 70%: Review these prerequisites first:

   • Linear algebra (matrix multiplication)

   • Basic calculus (derivatives)

   • Python programming

2. Score ≥ 70%: Proceed with Phase 5 content

3. Retake after Phase 5: Take the post-quiz to measure your progress

Remember: A low score is perfectly normal! This quiz shows where you’re starting, not where you’ll end up. 🚀