Classical ML
The algorithms that still power much of real-world machine learning. Drop points on a plane and watch clusters form, neighbours vote, and trees carve up the space.
Key concepts
Supervised Learning
Learn a mapping from inputs to known labels
Unsupervised Learning
Find structure, such as clusters, in data without labels
Decision Boundaries
Every classifier splits the input space into regions, one per class
Overfitting
A model that fits training noise does worse on new data
Visualizations
K-Means Clustering
BeginnerStep through assigning points to centroids and moving centroids to the mean
ClusteringCentroidsElbow MethodUnsupervised
Open K-Nearest Neighbors
BeginnerClassify a point by majority vote of its neighbours and see the decision boundary change with k
ClassificationDistance MetricsDecision BoundaryOverfitting
Open Decision Trees
IntermediateGrow a tree split by split using Gini impurity and see how it partitions the plane
Gini ImpurityInformation GainDepthOverfitting
Open Gradient Descent
IntermediateRoll down a loss surface and compare learning rates, momentum and Adam
Learning RateMomentumAdamLoss Surface
Open