Logging and Progress Control

nonconform uses Python logging for strategy-level progress and warnings. Aggregation progress bars are controlled by ConformalDetector(verbose=...).

Quick Setup

import logging

# Development: show progress/info
logging.getLogger("nonconform").setLevel(logging.INFO)

# Production: warnings and errors only
logging.getLogger("nonconform").setLevel(logging.WARNING)

verbose vs Logging Level

• verbose=True shows aggregation progress during compute_p_values() and score_samples().
• Logging level controls strategy-level progress output and warnings.

import logging
from sklearn.ensemble import IsolationForest

from nonconform import ConformalDetector, CrossValidation

logging.getLogger("nonconform").setLevel(logging.INFO)

detector = ConformalDetector(
    detector=IsolationForest(random_state=42),
    strategy=CrossValidation(k=5),
    score_polarity="auto",
    verbose=True,
    seed=42,
)
detector.fit(X_train)
_ = detector.compute_p_values(X_test)

Common Patterns

Quiet Production Output

import logging

logging.getLogger("nonconform").setLevel(logging.WARNING)

detector = ConformalDetector(
    detector=base_detector,
    strategy=strategy,
    verbose=False,
)

Keep Aggregation Bars, Hide Strategy Details

import logging

logging.getLogger("nonconform").setLevel(logging.WARNING)

detector = ConformalDetector(
    detector=base_detector,
    strategy=strategy,
    verbose=True,
)

Full Debugging

import logging

logging.basicConfig(
    level=logging.DEBUG,
    format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
)

Typical Progress Messages

Message	Component	Trigger
CV fold training (N folds)	Cross-validation strategy	fit()
Bootstrap training (N folds)	Bootstrap strategy	fit()
Aggregating N models	Ensemble aggregation	compute_p_values() / score_samples() with verbose=True

Logger Names

Useful logger namespaces:

• nonconform
• nonconform.resampling.*
• nonconform.weighting.*
• nonconform.fdr
• nonconform.adapters
• nonconform._internal.*

Configure specific namespaces if you want different levels per subsystem.

import logging

# Base level for package logs
logging.getLogger("nonconform").setLevel(logging.INFO)

# More detail for resampling strategies only
logging.getLogger("nonconform.resampling").setLevel(logging.DEBUG)