Question: Explain the rationale behind increasing model size to address high avoidable bias, including the potential negative side effects and how they can be mitigated.

Sample answer: Increasing the model size, such as adding more layers or neurons to a neural network, helps reduce high avoidable bias by allowing the model to better fit the training data. However, a larger model might increase variance and the risk of overfitting the training set. If this occurs, applying regularization techniques will usually eliminate the unwanted increase in variance.

Key points:

• Increasing model size reduces high avoidable bias.
• Larger models fit the training set better.
• A negative side effect is potential increased variance or overfitting.
• Regularization can be used to eliminate the variance increase.

Rubric: Full credit given for explaining why model size is increased, identifying the risk of increased variance/overfitting, and mentioning regularization as the solution.