Home▸Machine Learning & Neural Networks▸Supervised, Unsupervised and Reinforcement Learning: How Machines Actually Learn

🧠 Supervised, Unsupervised and Reinforcement Learning

Interactive 3D side-by-side simulation where feeding labeled data, unlabeled data or reward signals into three model panels shows each learning paradigm converging on a solution differently.

Machine Learning & Neural Networks3DModerate60 FPS

Three glowing platforms run the same raw data through supervised, unsupervised and reinforcement learning side by side, so you can watch a decision boundary, a set of cluster centroids and a reward-seeking agent all converge in real time.

🔬 What It Demonstrates

The supervised panel bends a boundary surface toward labeled points, the unsupervised panel drifts centroids toward the mean of nearby points with no labels at all, and the reinforcement panel lets an agent learn which grid path earns the most reward through trial and error.

🎮 How to Use

Adjust training speed, label noise, reward sparsity and dataset shape, then watch each panel's live accuracy, inertia and average reward stats update as the epoch counter climbs.

💡 Did You Know?

Reinforcement learning needs no labeled dataset at all — only a reward signal — which is why it's the paradigm behind game-playing agents like AlphaGo and most robotics control systems.

⚙ Under the hood

Interactive 3D side-by-side simulation where feeding labeled data, unlabeled data or reward signals into three model panels shows each learning paradigm converging on a solution differently.

supervised-learningunsupervised-learningreinforcement-learningai-fundamentalslearning-paradigmsmachine-learning

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install