Applying ML

Question 1

Should you use the training, validation, or test set of data to choose the final model?

Answer 1

You should use the validation set of data because it provides an independent evaluation of the final model’s performance

Question 2

Should you use the training, validation, or test set of data to report the accuracy of your model?

Answer 2

You should use the test set of data because it provides a more reliable assessment of how well your model performs on unseen data

Question 3

What does the training and testing error look like in a high bias model?

Answer 3

Both the training and test errors are HIGH

Question 4

What does the training and testing error look like in a high variance model?

Answer 4

The training error is LOW, and the testing error is HIGH

Question 5

What does the training and testing error look like in a model with both high bias and high variance

Answer 5

The training error is HIGH, and testing error is GREATER than the training error

Question 6

High Variance

Answer 6

Overfit

Question 7

High Bias

Answer 7

Underfit

Question 8

Can you use regularization in a neural network? What does it do?

Answer 8

YES, regularization can be used to prevent overfitting by using lambda

Question 9

Name 3 things to try to improve a model with high bias

Answer 9

1. Try getting additional features
2. Try adding polynomial features
3. Try decreasing lambda

Question 10

Name 3 things to try to improve a model with high variance

Answer 10

1. Get more training examples
2. Try smaller sets of features
3. Try increasing lambda

Question 11

Adding data to a model with high bias won’t help. Why?

Answer 11

This won’t help because high bias indicates the model is too simplistic and unable to capture the underlying patterns in the data, regardless of the amount of data available.

Question 12

Should you increase or decrease lambda used in regularization for a model with high variance?

Answer 12

Increase lambda to penalize overfitting