Lesson 4: Introduction to Neural Networks

Student Name:  

Date:  

Lab Partner(s):  

Class Period:  

Pre-Lab Questions

Before beginning the experiments, answer these questions based on your reading and class discussion:

1. Biological Inspiration

Question: What is a biological neuron and how does it work?

Answer:

Question: How are artificial neurons similar to biological neurons? How are they different?

Answer:

2. Network Architecture

Question: Draw and label a simple neural network with an input layer, one hidden layer, and an output layer.

Drawing space:

Labels to include:

• [ ] Input layer neurons
• [ ] Hidden layer neurons
• [ ] Output layer neurons
• [ ] Weights/connections
• [ ] Direction of information flow

3. Training Process

Question: In your own words, explain what happens during one training epoch.

Answer:

Part 1: Data Exploration

Dataset Information

Dataset used:  

Number of samples:  

Number of features:  

Number of classes:  

Class names:

1.  
2.  
3.  

Data Visualization Observations

Question: What patterns do you notice in the data visualizations?

Observations:

Question: Do the classes appear to be separable? Why or why not?

Answer:

Part 2: Network Architecture Design

Initial Network Design

Input layer size: _____ neurons (must match number of features)

Hidden layer(s):

• Layer 1: _____ neurons, Activation:  
• Layer 2: _____ neurons, Activation:   (if applicable)
• Layer 3: _____ neurons, Activation:   (if applicable)

Output layer size: _____ neurons (must match number of classes)

Output activation:  

Design Rationale

Question: Why did you choose this architecture?

Answer:

Part 3: Initial Training

Training Parameters

Learning rate:  

Number of epochs:  

Batch size (if applicable):  

Hypothesis

Before running: What do you predict will happen during training?

Prediction:

Initial Results

Final training accuracy:  

Final testing accuracy:  

Final training loss:  

Final testing loss:  

Training Curve Observations

Sketch or describe the loss curve:

Question: Did the loss decrease steadily? Were there any unusual patterns?

Observations:

Part 4: Experimentation

Experiment 1: Changing Number of Hidden Neurons

Hypothesis: How will changing the number of hidden neurons affect performance?

Prediction:

Hidden Neurons	Train Accuracy	Test Accuracy	Train Loss	Test Loss	Training Time
5
10
20
50
100

Observations:

Question: What is the optimal number of hidden neurons? Why?

Answer:

Experiment 2: Learning Rate Variation

Hypothesis: How will changing the learning rate affect training?

Prediction:

Learning Rate	Train Accuracy	Test Accuracy	Epochs to Converge	Notes
0.001
0.01
0.1
0.5
1.0

Observations:

Question: What happened when the learning rate was too high? Too low?

Answer:

Experiment 3: Number of Epochs

Hypothesis: How many epochs are needed for the network to converge?

Prediction:

Epochs	Train Accuracy	Test Accuracy	Overfit? (Y/N)
100
500
1000
2000

Observations:

Question: At what point did you observe overfitting (if any)?

Answer:

Experiment 4: Activation Functions

Hypothesis: How do different activation functions affect performance?

Prediction:

Activation (Hidden Layer)	Train Accuracy	Test Accuracy	Notes
ReLU
Sigmoid
Tanh

Observations:

Question: Which activation function worked best? Why do you think so?

Answer:

Part 5: Advanced Exploration (Optional)

Challenge: MNIST Digits

If you completed the MNIST challenge:

Network architecture used:

Input: _____ → Hidden: _____ → Output: _____

Training accuracy:  

Testing accuracy:  

Challenges encountered:

How you overcame them:

Part 6: Analysis and Reflection

Key Findings

Question: What were your three most important discoveries during this lab?

1.  

1.  

1.  

Connection to Real World

Question: Where do you see neural networks being used in your daily life?

Examples:

Question: What surprised you most about how neural networks learn?

Answer:

Challenges and Solutions

Question: What was the most challenging part of this lab?

Challenge:

Question: How did you overcome this challenge?

Solution:

Reflection Questions

1. Understanding Neural Networks

Question: Explain how a neural network learns to recognize patterns. Use an example from your experiments.

Answer:

2. Overfitting and Underfitting

Question: What is overfitting? Did you observe it in your experiments? How could you prevent it?

Answer:

3. Architecture Decisions

Question: If you were building a neural network to recognize handwritten digits, how would you design it? Explain your choices.

Architecture design:

Rationale:

4. Ethical Considerations

Question: Neural networks are used in facial recognition, hiring decisions, and criminal justice. What ethical concerns should we consider?

Concerns:

Question: How can we ensure neural networks are used responsibly?

Ideas:

5. Future Learning

Question: What would you like to learn next about neural networks or AI?

Topics of interest:

Summary and Conclusions

Main Takeaways

List 5 key concepts you learned:

1.  

1.  

1.  

1.  

1.  

Best Practices Discovered

What strategies worked well for training neural networks?

Areas for Improvement

What would you do differently next time?

Self-Assessment

Rate your understanding of each concept (1 = Don't understand, 5 = Fully understand):

Concept	Rating (1-5)	Notes
Biological inspiration for neural networks	_____
Network architecture components	_____
Forward propagation	_____
Backpropagation and gradient descent	_____
Activation functions	_____
Training process	_____
Loss functions	_____
Overfitting and underfitting	_____
Hyperparameter tuning	_____
Real-world applications	_____

Areas Where I Need More Help:

Teacher Feedback

Strengths:

Areas for Growth:

Grade:   / 50 points

Comments:

Additional Notes and Observations

Use this space for any additional thoughts, questions, or discoveries:

Code Snippets and Important Formulas

Record any important code or formulas here for future reference:

# Example: Network initialization

Key formulas:

Lab Completed: _____ / _____ / _____

Submitted: _____ / _____ / _____

Student Signature:  

Evolve AI Institute - Free AI Education for All

For questions or help: info@evolveaiinstitute.com