STUDY / 01

Explore Clustering

Put similar data into groups.

AlgorithmAssign points

Assign to centers, then update.

2clusters

0iteration

—SSE

Numbered markers: centers

Compare withExpand neighborhood

Connect dense neighborhoods; flag noise.

0clusters

0noise

Large dot: core · gray: noise / unvisited

Edit points with coordinates

Coordinates range from 0 to 1. You can also click the plot to add a point or remove a nearby point.

Point assignments

Point	x	y	K-means	DBSCAN
1	0.597	0.585	Unassigned	Unvisited
2	0.692	0.239	Unassigned	Unvisited
3	0.582	0.559	Unassigned	Unvisited
4	0.798	0.269	Unassigned	Unvisited
5	0.630	0.448	Unassigned	Unvisited
6	0.373	0.551	Unassigned	Unvisited
7	0.627	0.504	Unassigned	Unvisited
8	0.892	0.339	Unassigned	Unvisited
9	0.556	0.620	Unassigned	Unvisited
10	0.961	0.488	Unassigned	Unvisited
11	0.521	0.620	Unassigned	Unvisited
12	0.795	0.279	Unassigned	Unvisited
13	0.618	0.476	Unassigned	Unvisited
14	0.379	0.481	Unassigned	Unvisited
15	0.616	0.545	Unassigned	Unvisited
16	0.573	0.226	Unassigned	Unvisited
17	0.625	0.490	Unassigned	Unvisited
18	0.799	0.284	Unassigned	Unvisited
19	0.286	0.706	Unassigned	Unvisited
20	0.593	0.244	Unassigned	Unvisited
21	0.258	0.671	Unassigned	Unvisited
22	0.850	0.301	Unassigned	Unvisited
23	0.061	0.589	Unassigned	Unvisited
24	0.519	0.285	Unassigned	Unvisited
25	0.038	0.457	Unassigned	Unvisited
26	0.383	0.552	Unassigned	Unvisited
27	0.511	0.660	Unassigned	Unvisited
28	0.361	0.499	Unassigned	Unvisited
29	0.341	0.737	Unassigned	Unvisited
30	0.401	0.358	Unassigned	Unvisited
31	0.475	0.674	Unassigned	Unvisited
32	0.895	0.337	Unassigned	Unvisited
33	0.365	0.750	Unassigned	Unvisited
34	0.453	0.321	Unassigned	Unvisited
35	0.323	0.676	Unassigned	Unvisited
36	0.361	0.511	Unassigned	Unvisited
37	0.596	0.516	Unassigned	Unvisited
38	0.390	0.412	Unassigned	Unvisited
39	0.641	0.409	Unassigned	Unvisited
40	0.925	0.369	Unassigned	Unvisited
41	0.542	0.667	Unassigned	Unvisited
42	0.743	0.274	Unassigned	Unvisited
43	0.191	0.647	Unassigned	Unvisited
44	0.879	0.355	Unassigned	Unvisited
45	0.183	0.664	Unassigned	Unvisited
46	0.963	0.440	Unassigned	Unvisited
47	0.168	0.640	Unassigned	Unvisited
48	0.514	0.317	Unassigned	Unvisited
49	0.542	0.626	Unassigned	Unvisited
50	0.584	0.228	Unassigned	Unvisited
51	0.208	0.664	Unassigned	Unvisited
52	0.858	0.316	Unassigned	Unvisited
53	0.034	0.422	Unassigned	Unvisited
54	0.690	0.257	Unassigned	Unvisited
55	0.615	0.561	Unassigned	Unvisited
56	0.890	0.342	Unassigned	Unvisited
57	0.254	0.719	Unassigned	Unvisited
58	0.926	0.364	Unassigned	Unvisited
59	0.447	0.724	Unassigned	Unvisited
60	0.372	0.568	Unassigned	Unvisited
61	0.592	0.608	Unassigned	Unvisited
62	0.385	0.388	Unassigned	Unvisited
63	0.633	0.457	Unassigned	Unvisited
64	0.467	0.320	Unassigned	Unvisited
65	0.588	0.564	Unassigned	Unvisited
66	0.380	0.378	Unassigned	Unvisited
67	0.608	0.543	Unassigned	Unvisited
68	0.488	0.311	Unassigned	Unvisited
69	0.421	0.704	Unassigned	Unvisited
70	0.378	0.534	Unassigned	Unvisited
71	0.044	0.414	Unassigned	Unvisited
72	0.338	0.490	Unassigned	Unvisited
73	0.106	0.629	Unassigned	Unvisited
74	0.930	0.497	Unassigned	Unvisited
75	0.118	0.599	Unassigned	Unvisited
76	0.723	0.244	Unassigned	Unvisited
77	0.348	0.713	Unassigned	Unvisited
78	0.442	0.328	Unassigned	Unvisited
79	0.577	0.579	Unassigned	Unvisited
80	0.380	0.491	Unassigned	Unvisited
81	0.066	0.501	Unassigned	Unvisited
82	0.401	0.417	Unassigned	Unvisited
83	0.623	0.548	Unassigned	Unvisited
84	0.591	0.252	Unassigned	Unvisited
85	0.080	0.580	Unassigned	Unvisited
86	0.584	0.252	Unassigned	Unvisited
87	0.609	0.559	Unassigned	Unvisited
88	0.487	0.365	Unassigned	Unvisited
89	0.595	0.569	Unassigned	Unvisited
90	0.385	0.427	Unassigned	Unvisited
91	0.035	0.458	Unassigned	Unvisited
92	0.368	0.556	Unassigned	Unvisited
93	0.051	0.539	Unassigned	Unvisited
94	0.926	0.456	Unassigned	Unvisited
95	0.200	0.672	Unassigned	Unvisited
96	0.542	0.257	Unassigned	Unvisited
97	0.020	0.446	Unassigned	Unvisited
98	0.500	0.319	Unassigned	Unvisited
99	0.641	0.491	Unassigned	Unvisited
100	0.862	0.317	Unassigned	Unvisited
101	0.228	0.700	Unassigned	Unvisited
102	0.374	0.505	Unassigned	Unvisited
103	0.087	0.554	Unassigned	Unvisited
104	0.911	0.441	Unassigned	Unvisited
105	0.057	0.492	Unassigned	Unvisited
106	0.779	0.242	Unassigned	Unvisited
107	0.134	0.605	Unassigned	Unvisited
108	0.368	0.527	Unassigned	Unvisited
109	0.030	0.420	Unassigned	Unvisited
110	0.797	0.275	Unassigned	Unvisited
111	0.292	0.698	Unassigned	Unvisited
112	0.605	0.246	Unassigned	Unvisited
113	0.065	0.516	Unassigned	Unvisited
114	0.420	0.416	Unassigned	Unvisited
115	0.111	0.578	Unassigned	Unvisited
116	0.819	0.257	Unassigned	Unvisited
117	0.264	0.710	Unassigned	Unvisited
118	0.673	0.237	Unassigned	Unvisited
119	0.623	0.468	Unassigned	Unvisited
120	0.837	0.279	Unassigned	Unvisited

Clustering: methods and experiments ↗

Moons: compare k-means with DBSCAN or single linkage. Stretched clusters: compare spherical k-means with full-covariance Gaussian mixtures. Unequal density: test how a single DBSCAN radius behaves. Uniform noise: clustering can still produce groups where no natural groups exist.

Ward minimizes added within-cluster variance. Single linkage uses the nearest pair; complete uses the farthest; average uses mean pairwise distance. Gaussian mixtures use EM and soft memberships; plotted colors select the largest membership. Mean shift uses a flat neighborhood kernel and merges modes within one bandwidth. Radius and min-points stay fixed per run. One playback step means a different operation in each method. Clustering reference ↗

Source / Naftali Harris ↗