What is gradient descent in Keras training?

Updated May 16, 2026

Short answer

Gradient descent minimizes loss by updating weights iteratively.

Deep explanation

It computes gradients of loss function and updates weights in opposite direction.

Real-world example

Used in almost all neural network training tasks.

Common mistakes

  • Using too large learning rate.

Follow-up questions

  • What is stochastic gradient descent?
  • What is learning rate?

More Keras interview questions

View all →