Quiz: Loss Functions - How Neural Networks Measure Their Mistakes

Question 1 of 6

What is the primary purpose of a loss function in a neural network?

To measure how wrong the predictions are

To activate neurons in the network

To generate predictions from input data

To store the network's weights and biases

Question 2 of 6

What are the three steps in the flow that connects predictions to learning?

Forward propagation → Loss function → Backward propagation

Backward propagation → Loss function → Forward propagation

Loss function → Forward propagation → Backward propagation

Activation function → Loss function → Weight update

Question 3 of 6

What does MSE stand for and what type of problems is it used for?

Mean Squared Error, used for regression problems

Mean Squared Error, used for classification problems

Maximum Standard Error, used for regression problems

Mean Statistical Estimate, used for optimization problems

Question 4 of 6

Why are errors squared in the MSE loss function?

To make all errors positive and penalize big mistakes more

To make the calculation faster for computers

To convert percentages into whole numbers

To match the format of the activation functions

Question 5 of 6

In the cat classifier example with Binary Cross-Entropy, if the actual label is 1 (cat) and the network predicts 0.9, what is the approximate loss?

0.10

0.22

1.61

0.9

Question 6 of 6

Why shouldn't you use MSE for classification problems?

MSE treats all errors equally and doesn't penalize confident wrong predictions as effectively as cross-entropy

MSE is too slow for classification tasks

MSE only works with continuous numbers above 1000

MSE requires more memory than cross-entropy