Lesson 5: Building and Visualizing Neural Networks (1 hour)
Learning Objectives
- Use tools to build simple neural networks
- Visualize how networks process information
- Understand network architecture through hands-on exploration
- See real neural networks in action
Materials Needed
- Internet-connected devices
- Neural network playground/visualization tools
- Student notebooks
- Optional: Simple neural network building tools
Time Breakdown
- Review network types (5 min)
- Introduction to visualization tools (10 min)
- Hands-on: Neural Network Playground (25 min)
- Building a simple network (15 min)
- Wrap-up and unit review (5 min)
Activities
1. Review Network Types (5 min)
- What are the three main types we learned?
- When would you use each?
- Bridge: "Today we'll build and visualize networks ourselves"
2. Introduction to Visualization Tools (10 min)
Why Visualize?
- Neural networks are complex
- Visualizations help understand what's happening
- See how data flows, how networks learn
- Makes abstract concepts concrete
Tools We'll Use:
- Neural Network Playground (TensorFlow)
- Other online visualization tools
- Interactive demos
What We'll Do:
- Build simple networks
- Watch them learn
- See how architecture affects learning
- Experiment with different settings
3. Hands-On: Neural Network Playground (25 min)
Activity: Explore Neural Network Playground
If TensorFlow Playground is available, use it. Otherwise, use similar tool or simulation.
Part 1: Basic Network (10 min)
- Open Neural Network Playground
- Start with simple dataset (spiral or circles)
- Build network: 2 inputs, 1 hidden layer (4 neurons), 2 outputs
- Watch it learn
- Observe:
- How accuracy improves
- How decision boundary changes
- How weights adjust
Part 2: Experiment with Architecture (10 min)
- Try different numbers of hidden layers:
- 1 layer
- 2 layers
- 4 layers
- Observe: How does depth affect learning?
- Try different numbers of neurons:
- Few neurons (2-4)
- Many neurons (8-16)
- Observe: How does width affect learning?
Part 3: Experiment with Learning (5 min)
- Try different learning rates
- Observe: Too fast? Too slow? Just right?
- Try different activation functions (if available)
- Observe: How does it affect learning?
Key Questions to Answer:
- What architecture works best for this problem?
- How does depth vs. width affect learning?
- What happens with too many layers? Too few?
- How does learning rate affect training?
4. Building a Simple Network (15 min)
Activity: Design Your Network
Option 1: Using Visualization Tool
- Students design network for a specific problem
- Choose: Number of layers, neurons per layer
- Explain: Why did you choose this architecture?
- Test: Does it work? Why or why not?
Option 2: Physical Model
- Students create physical model of network
- Use paper, string, or craft materials
- Show: Layers, connections, data flow
- Label: Input, hidden, output layers
Option 3: Digital Diagram
- Students draw network architecture
- Choose problem (e.g., recognizing animals)
- Design: How many layers? Neurons? Why?
- Present to class
Reflection Questions:
- What architecture did you choose? Why?
- What would happen if you changed it?
- What did you learn about network design?
5. Wrap-Up and Unit Review (5 min)
Unit 3 Summary:
- Neural networks: Inspired by brains, layers of connected neurons
- Learning: Adjusting weights based on errors
- Deep learning: Many layers for complex patterns
- Types: Feedforward, CNN (images), RNN (sequences)
- Visualization: Tools help understand networks
Key Takeaways:
- Neural networks are powerful pattern learners
- Structure matters (depth, width, type)
- Learning is about adjusting weights
- Visualization helps understanding
Preview Unit 4: AI Applications - We'll explore how neural networks and other AI techniques are used in real-world applications like computer vision, language processing, and more!
Differentiation Strategies
- Younger students: Focus on visual exploration, simpler networks, guided activities
- Older students: Explore more complex architectures, analyze performance, research optimization
- Struggling learners: Use guided exploration, simpler tasks, more support
- Advanced learners: Research specific architectures, explore hyperparameter tuning, analyze training dynamics
Assessment
- Participation in hands-on activities
- Quality of network design
- Understanding demonstrated through observations
- Unit quiz completion
Unit 3 Assessment Rubric
Formative Assessment (Throughout Unit)
- Participation in discussions: 20%
- Hands-on activities: 30%
- Reflection journal entries: 20%
- Understanding checks: 30%
End-of-Unit Assessment
Unit Quiz (20 questions, open-note):
- Neural network basics (5 questions)
- Learning process (5 questions)
- Deep learning (5 questions)
- Network types (5 questions)
Project: Neural Network Visualization/Design
- Create visualization or diagram of neural network
- Explain: Structure, how it learns, what it could be used for
- Option 1: Digital diagram with explanations
- Option 2: Physical model with labels
- Option 3: Written description with architecture diagram
Rubric for Project
| Criteria | Excellent (4) | Good (3) | Satisfactory (2) | Needs Improvement (1) |
|---|---|---|---|---|
| Accuracy | Architecture is correct and well-designed | Architecture is mostly correct | Architecture has some errors | Architecture has significant errors |
| Understanding | Shows deep understanding of concepts | Shows good understanding | Shows basic understanding | Shows limited understanding |
| Explanation | Clear, detailed explanation of all components | Good explanation of most components | Basic explanation | Limited or unclear explanation |
| Creativity | Highly creative and well-executed | Creative and well-executed | Some creativity | Little creativity |
Unit 3 Resources
Required Tools
- Neural Network Playground: https://playground.tensorflow.org/ (or similar)
- Internet connection for visualizations
Recommended Exploration
- 3Blue1Brown Neural Networks video series (age-appropriate sections)
- Interactive neural network visualizations
- "How Neural Networks Work" animations
Teacher Notes
- Neural networks are abstract - visualization is crucial
- Students may struggle with concepts - use lots of analogies
- Hands-on exploration helps understanding
- Some students may want to dive deeper - have resources ready
- Be prepared for technical issues with online tools
Unit 3 Extension Activities (Optional)
For Advanced Students
- Research specific neural network architectures (ResNet, Transformer, etc.)
- Explore how backpropagation actually works mathematically
- Research different activation functions and when to use them
- Explore neural network optimization techniques
- Research current state-of-the-art neural networks
For Students Needing More Support
- Create visual flashcards of neural network components
- Draw diagrams showing different network types
- Make a simple guide: "Parts of a Neural Network"
- Practice identifying network types with examples
- Create physical models of simple networks
Next Unit Preview
Unit 4: AI Applications will explore how AI is used in the real world. We'll dive into computer vision (how AI sees), natural language processing (how AI understands language), speech recognition, and robotics. Get ready for hands-on experiments with real AI tools!