AIC and BIC: Choosing the Right Model Without Overfitting

Derive the log-likelihood from first principles, then build AIC and BIC as penalised model selection criteria. Compare polynomial fits, visualise the bias-variance tradeoff, and understand when to use each criterion.