Jun 13Vibe with Hermes Agent — Bengaluru · RSVP
ToolsMCPBlogResearchCommunityStar on GitHub
All Tools
E
DataFreeOpen Source

ELI5

Debug ML models with feature importance and prediction explanations

MIT

ABOUT

Machine learning models often behave unexpectedly — they rely on spurious correlations, overfit to noise, or fail silently on edge cases. Eli5 provides simple, visual tools for inspecting what a model has learned: it computes per-feature importance weights, highlights which tokens matter in text classifiers, shows decision trees in readable formats, and measures how prediction quality degrades when each feature is shuffled. This makes it easy to catch data leakage, confirm that important features are sensible, and communicate model behavior to non-experts.

INSTALL
pip install eli5

INTEGRATION GUIDE

1. Inspect feature weights in a linear classifier to catch unexpected spurious correlations in training data 2. Highlight the most influential tokens in a text classifier to verify it uses meaningful linguistic patterns 3. Debug an XGBoost model by comparing permutation importance against built-in feature importance metrics 4. Explain individual predictions with per-feature contribution breakdowns for auditability 5. Validate feature engineering by measuring which engineered features actually improve prediction quality

TAGS

pythonexplainabilitydebuggingfeature-importanceinspection