Evaluate your understanding of this unit

Use the Multiverse knowledge check to test how well you can recall the ideas, trade-offs, and techniques from this unit.

What is a common consequence of a model with high variance when deployed in a business environment?

A) The model performs consistently across all inputs.B) The model ignores relevant patterns in the data.C) The model generalises well to unseen data.D) The model behaves unpredictably when exposed to new data.

Which of the following best describes the tradeoff between model complexity and business value?

A) Increasing complexity always improves model performance and value.B) A simpler model will always outperform a complex one.C) The best model balances complexity to optimise accuracy, interpretability, and generalisation.D) Business value increases linearly with model complexity.

As model complexity increases, which of the following is most likely to happen?

A) Bias increases and variance decreases.B) Both bias and variance decrease.C) Bias decreases and variance increases.D) Both bias and variance increase.

Which scenario best illustrates underfitting due to high bias?

A) A model that performs perfectly on training data but poorly on validation data.B) A model that has low error on both training and validation datasets.C) A model that makes similar errors across all datasets and fails to capture patterns.D) A model that reacts strongly to changes in input data.

What does the "sweet spot" in the bias-variance tradeoff curve represent?

A) The point of highest model complexity.B) The lowest possible irreducible error.C) The point where bias and variance are both zero.D) The point where total error is minimised and generalisation is optimised.

Which technique is most appropriate for reducing overfitting?

A) Adding more features to the model.B) Decreasing model complexity.C) Ignoring validation performance during training.D) Using only training accuracy to guide tuning.

Which of the following changes is most likely to reduce underfitting?

A) Decreasing the number of model parameters.B) Using regularisation to shrink weights.C) Increasing model complexity to better capture data patterns.D) Removing features to avoid noise.

Which of the following is a strategy to improve model generalisation?

A) Train on a smaller dataset.B) Use ensemble methods like bagging.C) Maximise model depth regardless of validation results.D) Only tune the model using training error.

Which of the following scenarios best illustrates the use of regularisation to support fairness?

A) A model is trained faster by skipping irrelevant training records.B) A model's prediction accuracy improves for underrepresented groups after reducing reliance on zip code.C) A model adds new features to improve coverage of all population groups.D) A model learns to ignore missing values during training.

You notice that a credit risk model underpredicts repayment likelihood for applicants under 25. Which strategy would most directly help improve fairness in this case?

A) Remove all age-related features from the model.B) Apply L1 regularisation to increase model sparsity.C) Expand the training dataset to include more examples from younger applicants.D) Increase model complexity to capture more subtle patterns in the full dataset.

Action item: Knowledge check

Complete the quiz in the learning flow and use your results to identify which topics you want to revisit before moving on.

In progress