Round Groups Versus Moons — medium Machine Learning in Code problem | Incognition

Round Groups Versus Moons

10 pts · 10 coins

Machine Learning in CodeMedium

Round Groups Versus Moons

Two interlocking crescent moons are the classic shape that k-means gets wrong. Measure how wrong, and how DBSCAN does on the same points.

Task: write moon_scores(n_samples, noise, eps).

Make the data with make_moons(n_samples=n_samples, noise=noise, random_state=0). It returns X, y.
Fit KMeans(n_clusters=2, random_state=0, n_init=10) on X.
Fit DBSCAN(eps=eps, min_samples=5) on X.
Score each one against y with adjusted_rand_score.
Return [kmeans_score, dbscan_score], each as a plain float rounded to 3 decimal places.

Leave DBSCAN's noise points (-1) in the labels as they are. The score treats them as one more group.

Examples

n_samples = 200

noise = 0.05

eps = 0.3

returns [0.256, 1.0]

n_samples = 200

noise = 0.05

eps = 0.05

returns [0.256, 0.0]

Hints

solution.py

1from sklearn.cluster import DBSCAN, KMeans
2from sklearn.datasets import make_moons
3from sklearn.metrics import adjusted_rand_score
4
5
6def moon_scores(n_samples, noise, eps):
7    pass
8