Virtual Labs

Evaluation Metrics, Overfitting/Underfitting

What is the primary cause of overfitting in machine learning models?

a: Insufficient training data Explanation

Explanation

b: Using too few parameters in the model Explanation

Explanation

c: Learning noise instead of general patterns Explanation

Explanation

d: Using a very simple model Explanation

Explanation

Which of the following is a common symptom of underfitting?

a: High training accuracy but low test accuracy Explanation

Explanation

b: Low accuracy on both training and test data Explanation

Explanation

c: Model performing exceptionally well on unseen data Explanation

Explanation

d: A high number of parameters leading to complex decision boundaries Explanation

Explanation

Why is a train/test split necessary in machine learning?

a: To make the model memorize the test data Explanation

Explanation

b: To evaluate the model's performance on unseen data Explanation

Explanation

c: To prevent the model from learning any patterns Explanation

Explanation

d: To increase the training accuracy Explanation

Explanation

What is the primary difference between supervised and unsupervised learning?

a: Supervised learning requires labeled data, while unsupervised learning does not Explanation

Explanation

b: Unsupervised learning uses neural networks, while supervised learning does not Explanation

Explanation

c: Supervised learning is only used for regression tasks Explanation

Explanation

d: Unsupervised learning always provides better accuracy Explanation

Explanation

When a model performs very well on training data but poorly on new data, it is called:

a: Underfitting Explanation

Explanation

b: Overfitting Explanation

Explanation

c: Good generalization Explanation

Explanation

d: High bias Explanation

Explanation

Which of these is usually the best first step to fix overfitting?

a: Make the model much more complex Explanation

Explanation

b: Get more training data Explanation

Explanation

c: Remove all regularization Explanation

Explanation

d: Use a linear model only Explanation

Explanation