What is convergence in Gradient Descent?

Updated May 16, 2026

Short answer

Convergence occurs when updates become very small and the algorithm reaches a minimum.

Deep explanation

Gradient Descent converges when gradients approach zero or changes in cost function become negligible. It may converge to local or global minima depending on the function.

Real-world example

Model training stopping when validation loss stabilizes.

Common mistakes

  • Assuming convergence always means global optimum.

Follow-up questions

  • What affects convergence speed?
  • Can GD fail to converge?

More Gradient Descent interview questions

View all →