Better to view on desktop for better experience

Linear Algebra Calculus & Optimization Probability & Statistics Machine Learning

←Machine Learning

DBSCAN

Grow clusters from dense regions and flag outliers as noise.

UnsupervisedClustering

Sections

Select a section to explore

Watch DBSCAN learn

Follow each update in real time. Scrub or play to inspect the algorithm.

Step 1 of 8

● Solid: Core Points · ◌ Hollow: Border Points · ⨂ Red: Noise / Outliers

Step 1/8

ε (eps) = 8.5

3.5s/step

02 Interactive demo

CURRENT LOSS · OBJECTIVE

—

Training objective at current step

WHAT'S HAPPENING

1. The Raw Data: Any Shape + Outliers

We have 74 data points: a large sweeping hook on the left, a compact circle on the right, and 5 lonely outliers. K-Means struggles here because it assumes round spherical blobs and forces every outlier into a group. DBSCAN instead discovers groups naturally by looking for crowded neighborhoods.

\text{Parameters: } \varepsilon = 8.5 \text{ (circle radius)}, \quad \text{MinPts} = 4 \text{ (neighbors needed)}