StackedML
Practice
Labs
Questions
Models
Pricing
Sign in
Questions
/
Optimization
/
Gradient Methods
/
Stochastic gradient descent
← Previous
Next →
721.
SGD Fixed Learning Rate Issue
medium
SGD with a fixed learning rate often fails to converge precisely to a minimum. Why?
A
It continues to take steps of the same size even near the minimum, causing oscillation around rather than convergence to it
B
It produces biased gradient estimates near the minimum since single samples oversample the decision boundary region
C
It gradually increases the effective step size near the minimum since gradient magnitudes grow as the loss flattens
D
It becomes equivalent to random walk near the minimum since gradient estimates from single samples are too noisy
Sign in to verify your answer
← Back to Questions