Neural Net From Scratch

This is now a notebook-first project.

Start with learn_neural_nets.ipynb.

After that, move to tic_tac_toe_value_net.ipynb.

It is designed to be gentle and interactive:

• short explanations
• tiny coding tasks
• "stop and think" prompts
• simple self-checks
• no full end-to-end solution handed to you immediately

Best way to use it

1. Open the notebook.
2. Read one section at a time.
3. Do the task before asking AI.
4. Google syntax when you get stuck.
5. Only ask for a hint after you have tried something small.

Environment setup

From this folder:

python3 -m venv .venv
source .venv/bin/activate
python --version

If you want to run the notebook in Jupyter, install one notebook tool inside the virtual environment:

pip install notebook
jupyter notebook

If you prefer VS Code, you can also open the .ipynb file there directly.

Project files

• learn_neural_nets.ipynb: first guided notebook, gentle and beginner-friendly
• tic_tac_toe_value_net.ipynb: second guided notebook, harder and more mathematical, with exact tic-tac-toe labels and explicit side_to_move
• basics.py: tiny warmup script
• run_and_demo.py: perceptron demo
• nn_scratch/: plain Python starter code you can compare against later
• tests/: simple unittest checks

Suggested rule for you and your friend

For each notebook section:

1. One person types.
2. One person explains what each line should do.
3. Swap roles every 10 to 15 minutes.

After this notebook

Once you finish the first notebook, the next good step is:

1. Open the tic-tac-toe notebook.
2. Build a hidden layer.
3. Work through loss and gradients more carefully.
4. Later compare your plain Python version with NumPy.