A model learns from data
Linear regression · gradient descent
Explore week ↗English Русский
A COURSE IN IDEAS & EXPERIMENTS · 2026
Mathematical foundations, modern machine learning and experiments you can explain. Find the weekly plan, study the ideas, and connect them to practice.
A little intuition. A little mathematics. A lot of discovery.
YOUR ROUTE THROUGH THE COURSE
Linear regression · gradient descent
Explore week ↗
Logistic regression · SVMs · kernels
Explore week ↗
Decision trees · ensembles
Explore week ↗
Feedforward networks · autograd
Explore week ↗Slides ↗ Textbook ↗ Tutorial ↗ Lab — Unavailable / not assigned
Convolutions · residual learning
Explore week ↗
RNNs · Transformers
Explore week ↗Slides ↗ Textbook ↗ Tutorial ↗ Lab — Unavailable / not assigned
Likelihood · Bayes · the VAE idea
Explore week ↗
Weeks 01–07 · 90 minutes
Explore week ↗Fourier · wavelets · EEG · SKAB
Explore week ↗
PCA · t-SNE · autoencoders · VAE
Explore week ↗SimCLR · CLIP · representation learning
Explore week ↗U-Net · YOLO · SAM · uncertainty
Explore week ↗PINN · inverse problems
Explore week ↗Reinforcement learning · RLHF
Explore week ↗Student-selected investigation
Explore week ↗No matching weeks. Try another topic or choose All weeks.
Materials appear as they are released. A dash means unavailable or not assigned; planned weeks link to the syllabus.
ML is easy twice: first when you just use it, and again when you discover the roots from which it all grows.How we learn and work together ↗
Start with the weekly plan, follow the lecture and tutorial, then return to the textbook for the detailed explanation. In each experiment, ask what changed, how it was evaluated and what the evidence supports.
Dates, remaining institutional details and unresolved policy decisions are published separately. Items marked TODO(instructor) await instructor confirmation.