All problems
5730 problems across 15 categories, from vectors and gradients to backprop, attention, and Q-learning. Each one is a self-contained Python exercise with tests that run in your browser — no account needed.
Categories
Vectors, matrices, decompositions. The bedrock of every ML primitive.
Derivatives, gradients, Jacobians, Hessians. How things change.
Mean, variance, covariance, distributions. What the data tells you.
Bayes, expectation, counting. The language of uncertainty.
Regressions, trees, clustering, metrics. The classic toolkit, from scratch.
Activations, forward pass, backprop. The core of every neural net.
Tokenize, embed, compare. Text into numbers.
GD, momentum, Adam, schedules. How models actually learn.
Search, sort, graph traversal, DP. The timeless core of computer science.
Stacks, queues, trees, heaps, tries. The containers that make algorithms fast.
Convolutions, filters, pooling, edges. Images as numbers, from scratch.
MDPs, value iteration, Q-learning, policy gradients. Agents that learn by acting.
Autocorrelation, AR, MA, ARMA, forecasts. Data that moves through time.
PageRank, shortest paths, centrality. Networks, traversed.
Entropy, KL divergence, mutual information. How much does a message tell you?
Looking for something specific?
Use the filters above to search every problem by title, id, or category, or open the home page and press ⌘K to jump straight to one.
Search on the home page