What does the bias-variance trade-off describe in model development?

• A. High bias causes underfitting; high variance causes overfitting; balance model complexity, data quality, and training data size to achieve generalization.
• B. High bias causes overfitting; high variance causes underfitting.
• C. There is no trade-off; increasing model complexity always improves performance.
• D. Bias and variance are unrelated to model performance.

Answer: (A)

Balancing bias and variance is essential in model development. Bias comes from errors in the learning algorithm or assumptions that make the model too simple to capture the true patterns, leading to underfitting. Variance arises from sensitivity to the training data, causing the model to fit noise and perform poorly on new data, i.e., overfitting. Since making a model more complex reduces bias but increases variance, and simpler models increase bias but reduce variance, there is a trade-off. The goal is to achieve good generalization by tuning model complexity, data quality, and the amount of training data. This is why high bias leads to underfitting and high variance leads to overfitting, with the aim of balancing these factors to generalize well.