Know what you're missing in ML math.
Structured learning paths from foundations to frontier research. Every theorem with exact assumptions. Every concept linked to its prerequisites. Find your gaps and know what to study next.
Where Are You Starting From?
Pick the path that matches your background. Each one tells you what to study, in what order, and why it matters.
Math to ML
You know calculus, linear algebra, and some probability. You want to understand the theory behind gradient descent, generalization, and why models work.
For PractitionersEngineer to ML Theory
You can train models and ship code. You want to understand why things work, when they break, and what the theorems actually say.
Bridge PathStats to Deep Learning
You have a statistics background. You want to connect classical stats to modern deep learning: from MLE to transformers.
Interview PrepML Research Interview Prep
Interview-oriented coverage: concentration, uniform convergence, optimization, attention, scaling laws. Exact theorem statements and failure modes.
The Concept Map
Click any node to start reading. Hover to see prerequisites.
Start With These
The most important concepts, explained with interactive diagrams and exact theorem statements.
Concentration Inequalities
Markov, Chebyshev, Hoeffding, Bernstein
Bias-Variance Tradeoff
The U-curve, when it breaks, double descent
Transformer Architecture
Self-attention, FFN, residual stream
Scaling Laws
Kaplan, Chinchilla, compute allocation
Grokking
Delayed generalization, phase transitions
Induction Heads
The circuit behind in-context learning
See It, Don't Just Read It
Interactive demos that show the mechanics directly.
Optimizer Race
SGD, momentum, RMSProp, and Adam on convex and non-convex landscapes. Switch loss surfaces and watch the trajectories.
All Interactive Demos
Bias-variance slider, softmax temperature, dropout masking, likelihood surfaces, activation functions, and more. All run in your browser.
Find Your Gap
A short diagnostic across probability, optimization, and learning theory. Points you at the topics you should read next.
How This Works
Prerequisite Chains
Every concept traces down to its foundations. The sidebar shows you what you need to know first. No concept floats without grounding.
Exact Theorems
Formal statements with explicit assumptions, proof sketches, failure modes, and when each result actually matters in practice.
Interactive Diagrams
Gradient descent contour plots, dropout masking, softmax temperature sliders, and attention heatmaps. Concepts you can manipulate, not just read.