← All challenges
Track · 7 daysLive

7 Days of ML

Load data, train models, and ship a real classifier — one focused build a day for a week.

  1. Day 1Your first modelBeginnerLoad a real dataset, train a classifier, and measure how well it does on data it has never seen.~30 min · 3 blanks→
  2. Day 2Recognise handwritten digitsBeginnerTrain a k-nearest-neighbours model to read handwritten digits from 8x8 pixel images.~30 min · 2 blanks→
  3. Day 3Predict a numberIntermediateSwitch from categories to continuous values: fit a linear regression that predicts disease progression.~30 min · 2 blanks→
  4. Day 4Bend the boundaryIntermediateTwo interleaving moons no straight line can split. Reach for a model whose boundary can curve.~30 min · 1 blanks→
  5. Day 5Don't overfitIntermediateA deep tree can memorise its training data and still fail on new data. Use cross-validation to measure honestly.~30 min · 2 blanks→
  6. Day 6Scale it upAdvancedFeatures on wildly different scales wreck distance-based models. Fix it the right way with a pipeline.~30 min · 1 blanks→
  7. Day 7Ship a real classifierAdvancedThe capstone: train a Random Forest on real medical data, evaluate it, and read its confusion matrix.~30 min · 3 blanks→