Back to course
Next page
Interactive Quiz
Test your knowledge!
1
What is the main problem that regularization aims to solve in neural network training?
A
Underfitting due to a lack of data
B
Overfitting, meaning an overly precise fit to the training data
C
Slow convergence during optimization
D
Poor weight initialization
2
What function is added to the loss during L2 regularization?
A
The sum of absolute weights multiplied by a coefficient
B
The sum of squared weights multiplied by a coefficient
C
The sum of squared gradients of the weights
D
The sum of activations multiplied by a coefficient
3
What is the intuitive effect of limiting weight values with L2 regularization on the loss function?
A
It makes the loss function flatter, reducing abrupt variations
B
It increases the slope of the loss function for faster learning
C
It makes the loss function non-convex
D
It reduces the number of local minima in the loss function
4
What does the dropout method involve during neural network training?
A
Adding Gaussian noise to the weights at each iteration
B
Randomly setting certain weights to zero during training
C
Randomly setting certain activations to zero at each training step
D
Reducing the size of the network by removing certain layers
5
Why does dropout help reduce overfitting?
A
It increases the network’s learning capacity by adding neurons
B
It forces neurons to cooperate and prevents some neurons from relying on specific details
C
It speeds up convergence by decreasing the size of the training data
D
It replaces the loss function with a more robust function
Score: 0/5
Score: 0/5