☀️ Daily Morning Briefing

• Bias is error from overly simplistic assumptions (underfitting). Variance is sensitivity to noise in training data (overfitting).
• Total Error = Bias² + Variance + Irreducible Noise. As model complexity increases, bias ↓ but variance ↑.
• Reduce variance: regularization (L1/L2), dropout, early stopping, cross-validation, bagging/random forests.
• Reduce bias: more complex models, more features, less regularization, boosting methods.
• Diagnose with learning curves: if train ≈ val ≈ high → high bias; if train ≪ val → high variance.

from sklearn.model_selection import learning_curve
train_sizes, train_scores, val_scores = learning_curve(
    model, X, y, cv=5, scoring='neg_mean_squared_error'
)
# Plot mean train vs val score across training sizes
# Gap = variance; both high = bias