Model Building & Algorithm Learning

Decision Boundary Visualiser

Train different classifiers on the same two-feature problem, then inspect how their assumptions reshape the class boundary.

7 · Regression 8 · Clustering 9 · Decision boundaries 10 · Neural networks 11 · Gradient descent 12 · Ensembles 13 · Step-by-step

Lab concept guide

What to observe while you experiment

A decision boundary is the set of feature values where a classifier changes its decision. Its shape reflects both the data and the model family: linear models produce linear separators in the represented feature space, KNN is local, and trees form axis-aligned regions.

MechanismKeep the two-feature dataset fixed and train different classifiers so the only changed factor is the model’s decision rule.

Failure modeChoosing the most complex-looking boundary from the training plot without checking held-out performance or sensitivity to noise.

VerificationProbe points on both sides of a visible boundary and confirm their predictions/probabilities; compare training and validation behaviour after adding noise.

Experiment deliberately

Use one nonlinear dataset. Predict the boundary shape for logistic regression, KNN and a tree, then train and explain each difference.

DatasetClassifierComplexity / k5Noise0.12Seed

x y Click an existing point in the plot to remove it.

Untrained.

Feature space

The shaded surface is the predicted probability; the dark contour is the 0.5 decision boundary.