M1 · Linear Algebra
Linear Algebra and Matrix Mechanics for Machine Learning
Vector spaces, matrix representations, and geometric transformations.
Math, code, classical ML, deep learning, modern AI, then contest practice. Every prerequisite lands before you need it.
Problems, notebooks, and mock contests written at real USAAIO difficulty, not a generic ML blog.
Made by students, for students. No paywall, no course upsell. Just the path.
Six tracks, foundations through practice. Pick a track to see its modules.
Track 01
The linear algebra, calculus, probability, and optimization every model rests on.
M1 · Linear Algebra
Vector spaces, matrix representations, and geometric transformations.
M2 · Matrix Decompositions
Factorizing matrices into fundamental components: SVD, eigendecomposition, and friends.
M3 · Matrix Calculus
Gradients, derivatives, and optimization over matrix structures.
M4 · Optimization
Iterative algorithms and frameworks for finding optimal solutions.
M5 · Probability
Uncertainty, random variables, and probability distributions.
M6 · Information Theory
Entropy, information, and statistical divergence.
M7 · Probabilistic Models
Data distributions, variational inference, and generative architectures.
M8 · Stochastic Dynamics
Continuous-time processes, differential equations, and dynamic sampling.
bonus-rl · Reinforcement Learning
Sequential decision-making and dynamic programming. Not on the USAAIO syllabus.
Every panel below is real guide code on sample data. Type in the editor. Run it. Filter the problem set.
01 · training
02 · lessons
03 · problem set
04 · dashboard
We are students preparing for the same contest you are. There was no single place that taught the material in order, so we started writing one as we worked through it. Every module here is something we needed and could not find.
That is also why it stays free. Nobody is being paid, nothing is being upsold, and there is no tier of the guide you have to buy your way into. If it helps one more person qualify, it has done its job.
We learn each topic, then write the explanation we wish we had started with. When a derivation loses us, it gets rewritten until it does not.
Nothing here is locked away. Spot a rough explanation or a missing step, tell us, and it gets fixed. The guide is shaped by the people working through it.
New modules and problem sets go up as we finish them. You will see the guide grow rather than arrive fully formed.
Corrections and new lessons are welcome from anyone.
Yes. Every lesson, problem, notebook, and tool on the guide is free to use, with no account wall in front of the reading and no paid tier waiting behind it. We are students writing this for other students, so there is nothing to upsell and no plan to change that.
Not yet. We are still writing the first wave of lessons and problems, and we would rather ship something you can work through than a half-finished shell. Join the waitlist and you get one email the day the guide opens. No drip sequence, no weekly digest.
Comfortable high-school algebra is enough to start. The math track builds the linear algebra, calculus, probability, and optimization that later modules assume, and the Python track covers NumPy before you touch real models. If you already know some of that, you can skip ahead; the syllabus is ordered, not gated.
No. USAAIO Guide is an independent, community-built roadmap written by students preparing for the same contest. It is not affiliated with, endorsed by, or officially connected to USAAIO, the IOAI, the IAIO, or the Beaver-Edge AI Institute. For official rules, eligibility, and registration, go to usaaio.org.
No. Most of the Python in the guide runs in your browser through Pyodide, so a normal laptop is enough for the early tracks. Anything that actually needs a GPU (training a small network, running a heavier notebook) opens in Google Colab instead of asking you to set up CUDA locally.
Please do. Before launch, the most useful help is usually a clearer explanation, a missing prerequisite, or a problem that finally clicked for you. Come say hello on our Discord and tell us what tripped you up. Typo fixes count.