machine-learning-introduction

Shows the ML pipeline as three blocky panels: a dataset (with labels y in supervised mode, no labels in unsupervised), a parameterized model f(x; θ) with learnable parameters, and an empirical risk R(θ) that training minimizes. Animation cycles between supervised and unsupervised, and within each mode it shows θ updating over time while the risk decreases.

canvasclick to interact

t=0s

practical uses

01.Explain what a parameterized hypothesis f(x;θ) means before introducing specific model families (linear models, neural nets).
02.Motivate training as optimization by visualizing empirical risk decreasing with parameter updates.
03.Clarify the core distinction between supervised and unsupervised learning by showing how labels y change the objective.

technical notes

Pure Canvas2D, green-on-black retro grid with snapped coordinates for a blocky feel. Time-based animation uses (t/duration)%1 cycles: mode toggles every 3.6s and training progress runs on a 2.6s loop with cubic easing via ease(). Deterministic pseudo-random point placement uses a sine hash for stable visuals without external state.

← linear-regression minimum-spanning-trees →