Gradient Descent — Concept Check
Questions on Gradient Descent. Gradient descent is the loop that actually trains models: compute the gradient of the loss, take a small step downhill, repeat — with the learning rate controlling the step size.
Questions in this quiz
8 questions on Gradient Descent — Concept Check. Work through them here, then start the quiz to answer them and see your score.
1.A model's training process is taking too long to converge. What is a likely reason?
- ALearning rate is too small
- BUsing mini-batch gradient descent
- CLearning rate is too large
- DUsing stochastic gradient descent
2.During training, a model's loss suddenly increases after initially decreasing. What might be the cause?
- ALearning rate is too small
- BUsing Adam optimizer
- CUsing batch gradient descent
- DLearning rate is too large
3.Which of the following is true about mini-batch gradient descent?
- AIt is slower than batch gradient descent
- BIt uses only one example per step
- CIt balances the speed and accuracy of training
- DIt uses all data per step
4.A model diverges during training when using a learning rate of 1.1. What adjustment should be made?
- ADecrease the learning rate
- BSwitch to batch gradient descent
- CIncrease the learning rate
- DSwitch to stochastic gradient descent
5.What happens if the learning rate is set too high during gradient descent?
- AThe model will take longer to converge
- BThe model will stop learning
- CThe model may diverge
- DThe model will converge faster
6.A model trained with batch gradient descent is slow. What could be a solution?
- ASwitch to mini-batch gradient descent
- BIncrease the learning rate
- CDecrease the learning rate
- DSwitch to stochastic gradient descent
7.Which optimizer is essentially gradient descent with additional step rules?
- AAdam
- BBatch gradient descent
- CMini-batch gradient descent
- DStochastic gradient descent
8.What is the primary role of the learning rate in gradient descent?
- ASelects the data for training
- BAdjusts the loss function
- CControls the size of each step
- DDetermines the model architecture