ebonica-daml 0.2.0

Disagreement-Aware Multi-Learning (DAML) - A production-level Python ML library.

DAML: Disagreement-Aware Multi-Learning

DAML is a production-level Python machine learning library designed to detect ambiguous or difficult examples through ensemble variance, filtering them based on an adaptive disagreement threshold.

Key Features

• DAMLRegressor & DAMLClassifier: Core ML models compatible with the scikit-learn API. They dynamically retrain on difficult data points to improve performance.
• Drift Detection: Detect dataset feature shifts gracefully in production using simple thresholding rules matching your feature space.
• Hyperparameter Tuning: Cross-validation wrapper built specifically for fine-tuning the base ensemble.
• Scikit-Learn API: Seamless fit, predict, and partial_fit API style integrations.

Installation

pip install ebonica-daml

Or from source for development:

git clone https://github.com/yourusername/ebonica-daml.git
cd ebonica-daml
pip install -r requirements.txt
pip install -e .

Basic Usage

from daml import DAMLRegressor
from sklearn.datasets import make_regression

X, y = make_regression(n_samples=500, n_features=10)
model = DAMLRegressor(n_models=5, lambda_param=1.0)
model.fit(X, y)
predictions = model.predict(X)

See example_usage.py for a larger, complete end-to-end example dealing with hyperparameter tuning, model training, and data drift detection scenarios.