StackedML
Practice
Labs
Questions
Models
Pricing
Sign in
Questions
/
Optimization
/
Gradient Methods
/
Convergence
← Previous
Next →
139.
Convergence Criterion
easy
Which of the following is a common criterion used to detect convergence during training?
A
The training loss reaching a value lower than the validation loss by a fixed margin across all epochs
B
The validation accuracy exceeding the training accuracy indicating the model has generalized successfully
C
The number of training iterations exceeding a predefined maximum regardless of loss behavior
D
The gradient norm falling below a small threshold indicating parameters are near a stationary point
Sign in to verify your answer
← Back to Questions