Model Anatomy · Episode 01 · 30-second cut · 0:35
Gradient descent explained: how every AI model learns by walking downhill
Measure the slope, take a small step, repeat. In 30 steps the loss falls from 20.8 to almost zero, and the same idea trains neural networks and large language models.
What you’ll needPython 3.10+, NumPy
What you’ll need
- python.org/downloads →
Python 3.10 or newerRuns the code. Check yours with python3 --version. - code.visualstudio.com →
VS Code or CursorEither works; so does any editor you like. - git-scm.com/downloads →
gitGets the code from GitHub. 
NumPy 1.26+Math on many numbers at oncepip install "numpy>=1.26"
Set it up
- 1Get the code
git clone https://github.com/DayanEbrar0X/data-anatomy.ai.git cd data-anatomy.ai
- 2Make a virtual environmentIt keeps this project's packages separate from the rest of your computer.
python3 -m venv .venv source .venv/bin/activate
- 3Install the packagesrequirements.txt installs every lesson's packages. This one needs only: NumPy.
pip install -r requirements.txt # or just this lesson: pip install "numpy>=1.26"
- 4Run it
cd machine-learning/01-gradient-descent python3 src/descent.py
Project structure01-gradient-descent/
Project structure
machine-learning/01-gradient-descent/Open on GitHub →
- src/
- descent.pymainthe 8 lines from the video
- README.mdthe idea, how to run it, a line-by-line walkthrough, things to try
machine-learning/01-gradient-descent/src/descent.pyOpen on GitHub →
import numpy as np
loss = lambda w: .5 * (w[0]**2 + 10 * w[1]**2)
grad = lambda w: np.array([w[0], 10 * w[1]])
w = np.array([-4.0, 1.6]) # start high
lr, start = 0.17, loss(w) # step size
for step in range(30):
w = w - lr * grad(w) # step downhill
print(f"{start:.1f} -> {loss(w):.4f}")Transcript
This is how every AI on the planet learns. Picture it standing on a hill. Blindfolded. The hill is its mistakes.
Up high? Way off. Down at the bottom? Basically perfect.
It can't see the bottom, so it feels the slope under its feet, and takes one small step downhill. Then again. And again. Thirty tiny steps.
Its mistakes drop from 20.8 to almost zero. That's gradient descent. Feel the slope, step down, repeat.
Read the lesson on GitHub →- #gradientdescent
- #machinelearning
- #deeplearning
- #python
