HumachLearnLearn · explore · apply
Start learning
Working playground

Classification Playground

Generate different classification problems, train real browser-side algorithms step by step, and inspect the boundary and decision metrics.

Lab concept guide

What to observe while you experiment

A classifier learns a rule that maps feature values to class predictions or probabilities. Different algorithms impose different geometry on that rule, so the same dataset can produce linear, local, piecewise or ensemble decision boundaries.

MechanismGenerate a dataset, fit one model, and relate the visible boundary to the model assumption and the training points that constrain it.
Failure modeJudging a boundary only by training fit or assuming a more intricate boundary is automatically better.
VerificationCompare training and held-out decision metrics, then inspect several points near the boundary and confirm why their predicted class changes.
Experiment deliberately
Use the same dataset with logistic regression, KNN and a tree. Predict which boundary will be smooth, local or axis-aligned before training.
Ready.