Project 2 Midterm Prep Quiz

When the loss function becomes exactly zero

When the learning rate becomes constant

When the optimization process stops improving significantly

When weights increase indefinitely

The lowest value in a local region of the loss function

The average of all local minimums

The highest point on the cost curve

The absolute lowest value of the loss function across all parameter space

A temporary value reached before the global minimum

A point where the gradient is zero, but not the absolute minimum

A saddle point

The lowest value that gradient descent can never reach

It adds the absolute value of coefficients as a penalty term to the loss function.

It adds the square of coefficients as a penalty term to the loss function.

It reduces the learning rate during training.

It increases the number of features in the model.

To reduce training time

To increase model complexity

To reduce overfitting

To eliminate outliers from the dataset

Lasso

Ridge

Elastic Net

None of the above

A method that switches between L1 and L2 during training

A combination of L1 and L2 regularization

A regularization that focuses on maximizing entropy

A method used to increase recall

The highest accuracy achieved

The dividing line between different predicted classes

The level of regularization applied

The convergence point of gradient descent

The threshold value

The dividing line between different predicted classes

The confusion matrix

The ROC curve

(TP + FN) / (TP + FP + TN + FN)

(TP + TN) / (TP + TN + FP + FN)

TP / (TP + FP)

TP / (TP + FN)