Guided Micrograd Exercise

Use the original micrograd code as your base reference. Then complete these tasks.

Exercise A: Add ReLU

1. Add a relu() method to the Value class.
2. During the forward pass, output 0 when the input is negative and the input itself otherwise.
3. During the backward pass:
   • pass the gradient through unchanged if the input was positive
   • pass 0 backward if the input was negative

Write a quick test case:

• input -2.0 should produce output 0
• input 3.0 should produce output 3.0

Exercise B: Trace A Tiny Graph

Create values:

• a = 2.0
• b = -3.0
• c = 10.0
• d = a * b + c

Then:

1. Draw the graph.
2. Predict the sign of each gradient before calling backward.
3. Check your prediction after running backward.

Exercise C: Explain A Gradient In Plain Language

Pick one gradient from your graph and finish this sentence:

“If I changed this value upward a tiny bit, the final output would likely ______ because ______.”

Deliverable

Write a short note with:

• what you changed
• what worked
• what confused you
• one thing that feels clearer now