Which issue occurs when the learning algorithm fails to extrapolate a functional set of parameters from the training data?

Which issue occurs when the learning algorithm fails to extrapolate a functional set of parameters from the training data?

• A. Underfitting
• B. Overfitting
• C. Hyperparameters
• D. Accuracy

Answer: (A)

Underfitting happens when the model doesn’t have enough capacity or learning power to capture the underlying patterns in the data, so it can’t find a useful set of parameters that fit even the training data well. In this situation, the model’s representation is too simple or overly constrained, leading to high error on the training set and poor generalization. To address this, you can increase model complexity (more features or a more expressive model), reduce overly strong regularization, or provide more informative features or data so the model can learn the true relationships. By contrast, overfitting is when the model learns the training data too well, including noise, and then fails to generalize; hyperparameters are tunable settings, and accuracy is a performance metric, not the root cause described.

Underfitting happens when the model doesn’t have enough capacity or learning power to capture the underlying patterns in the data, so it can’t find a useful set of parameters that fit even the training data well. In this situation, the model’s representation is too simple or overly constrained, leading to high error on the training set and poor generalization. To address this, you can increase model complexity (more features or a more expressive model), reduce overly strong regularization, or provide more informative features or data so the model can learn the true relationships. By contrast, overfitting is when the model learns the training data too well, including noise, and then fails to generalize; hyperparameters are tunable settings, and accuracy is a performance metric, not the root cause described.