Before You Begin
Think of it like this: you wouldn't hand someone a race car and say "good luck" if they'd never driven a go-kart. This course is the go-kart. By the end, you'll understand every piece well enough to actually start building the real thing — not just nod along while someone else explains it.
Who This Is For
Two people, same starting line: an MBO student who can already program but hasn't touched EEG, sleep science, or machine learning yet — and the course's own founder, learning the same material alongside them. Neither of you needs to be an expert going in. You both need to come out the other side able to read, evaluate, and eventually extend real sleep-staging code.
Where This Is Headed
The real target: Dreamento, a real, open-source toolbox for EEG-based sleep and lucid-dream research, built around the ZMax headband. Its current built-in autoscorer isn't a neural network — it extracts handcrafted signal features and classifies them with LightGBM, a classical machine-learning method. The CNN/LSTM approach (the kind of neural network you've probably heard of) lives in the wider research literature this course builds toward — and, as you'll find in Part 2, a version of it is already sitting inside Dreamento's own codebase, just written in old, frozen-in-time code.
No EEG headband needed to start. Every lab in this course uses free, public sleep-recording data.
How The Course Is Built
- Part 1 — EEG & Sleep Fundamentals. What the raw signal actually is, how sleep is staged, and how to see it in the frequency domain.
- Part 2 — The Machine Learning Pipeline. A classical baseline, deep learning fundamentals, rebuilding a real sleep-staging network, and connecting it back to Dreamento's actual code.
- Part 3 — Lucid Dreaming Specifically. Sleep staging tells you when someone is in REM. It doesn't tell you they're lucid — that's a separate, harder, genuinely unsettled question, and Part 3 is where the honesty rules from the main Lucid Dreams course apply just as strictly to the research.
Every chapter follows the same shape: Learn the concept, Do a real hands-on lab (most run free in Google Colab, no install needed), and a clear Done when checkpoint so you know when you've actually got it, not just read about it.
DONE LOOKS LIKE
Not a finished product. A real, written technical proposal: which ML approach makes sense to actually pursue, and why — grounded in having built and compared the alternatives yourself, not guessed from a blog post.
Related
This course assumes you've already read the neuroscience modules in the main Lucid Dreams course — that's the "why," this is the "how to build it."