Contests

Six easy openers from across the curriculum — matrices, stats, probability, text, information, and pixels. Find your rhythm and bank a quick score.

Matrix products, norms, projections, and determinants, finishing with an adjugate and a QR solve. Ten problems, thirty minutes.

Factorials, combinations, permutations, and expectation, then Bayes against the clock. Short set, sharp thinking.

Mean, variance, median, and mode, finishing on covariance. Five descriptive-statistics problems, one sitting.

Sigmoid, entropy, Gini, KNN, and metrics, then three hard classics — k-means, tree splits, and the normal equation. The whole intro toolkit in one run.

ReLU and friends, a numerically stable softmax, an MLP forward pass, a conv layer, and a backprop gradient. The set ramps up fast.

A cross-category sampler from the full catalog: five easy openers, then five mediums spanning ML, RL, time series, vision, and optimization.

Ten genuinely hard problems — Jacobians, k-means, tree splits, backprop, and Adam, plus hard cuts from vision, RL, forecasting, graphs, and information theory. No warm-up.

From RGB plumbing to interpolation: eight computer-vision problems, ending on bilinear sampling. Pixels in, numbers out.

Policy evaluation, extraction, and improvement, building to value iteration to convergence. Eight reinforcement-learning problems.

Differences, lags, and rolling statistics, capped by a Holt-Winters step. Eight time-series problems against the clock.

Build the graph, traverse it, then hunt cycles and bridges. Eight graph-algorithms problems from adjacency to bridges.