evo-mars-grid-search

SkillSearch

Parallel grid search over DBSCAN hyperparameters for Mars cloud clustering. Aggregates per-image F1 and delta scores with proper averaging semantics.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the evo-mars-grid-search skill

What this skill tells your AI

The instructions your AI receives, as published by openlair/openskill in tasks-evolved/mars-clouds-clustering/environment/skills/evo-mars-grid-search/SKILL.md and read by ahel’s review.

Parallel grid search over DBSCAN hyperparameters.

Key Functions

  • load_data(citsci_path, expert_path) - Load CSVs, return grouped dicts
  • evaluate_hyperparameter_combo(ms, eps, sw, images, citsci_dict, expert_dict) - Eval one combo
  • run_grid_search(unique_images, citsci_dict, expert_dict, n_jobs) - Full parallel grid search
  • filter_results(results_df, min_f1=0.5) - Filter to F1 > 0.5

Usage

import sys
sys.path.insert(0, '/app/environment/skills/evo-mars-grid-search/scripts')
from utils import load_data, run_grid_search, filter_results

_, _, images, citsci_dict, expert_dict = load_data('citsci.csv', 'expert.csv')
results = run_grid_search(images, citsci_dict, expert_dict)
filtered = filter_results(results)

Averaging Rules

  • Loop over ALL unique images from expert dataset
  • Images with no citsci points: F1=0.0, delta=NaN
  • F1 average includes all zeros
  • Delta average excludes NaN values
  • Filter: only keep F1 > 0.5

Signals

GitHub stars
89
Forks
4
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
evo-mars-grid-search
Source
github.com/openlair/openskill