0 of 3 done
How to know whether a model is actually good — the discipline that separates a real result from a number that looks nice on training data. Part of the ML Engineer track.
- Evaluation Metrics — [Intermediate] — accuracy/precision/recall/F1, confusion matrices, regression metrics, and choosing the metric that matches the cost of each error
- Overfitting & Regularization — [Intermediate] — the bias-variance tradeoff and how to stop a model memorizing
- Validation & Tuning — [Intermediate → Advanced] — cross-validation (k-fold/LOOCV), ROC-AUC/log-loss, model selection, and hyperparameter tuning
Related
- Train/Val/Test Splits — the foundation honest evaluation rests on
- Classical ML — the models being evaluated
Jump to a subject
Software engineering
Infrastructure & security
Computer science
Hardware & engineering