Mars clustering

Skill cxcscmu/SkillLearnBench/skills/b4-skill-creator-gemini-3.1-flash-lite-preview/dbscan-parameter-tuning/mars-clustering

How to optimize DBSCAN hyperparameters for Mars cloud clustering. Use this skill whenever performing grid search, DBSCAN clustering, or evaluating F1 and delta for Mars datasets.From its SKILL.md

Install

npx -y skills add cxcscmu/SkillLearnBench --skill mars-clustering

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

SKILL.md

1.1 KB, 314 tokens by cl100k_base, as published. Nobody here has run it

Mars Cloud Clustering Optimization Workflow

1. Data Preparation

Load citsci_train.csv and expert_train.csv from /root/data/.
Group by file_rad to match annotations.

2. DBSCAN Custom Distance Metric

Implement the custom distance function: d(a, b) = sqrt((w * Δx)² + ((2 - w) * Δy)²)

3. Evaluation Metrics

F1 Score: Harmonic mean of Precision and Recall.
Delta: Standard Euclidean distance between matched centroids and expert points.
Greedy Matching: Match cluster centroids to expert points with max distance 100px.

4. Grid Search Parameters

min_samples: [3, 4, 5, 6, 7, 8, 9]
epsilon: [4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24]
shape_weight: [0.9, 1.0, 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9]

5. Output

Filter by mean F1 > 0.5.
Identify Pareto-optimal frontier.
Save to /root/pareto_frontier.csv.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.