Question

What are the main benefits of Mini-batch Gradient Descent?

Answer 1

Below are the main benefits of mini batch gradient descent :

1. It is computationally efficient compared to stochastic gradient descent.
2. It improves generalization by finding flat minima.
3. It improves convergence by using mini-batches. We can approximate the gradient of the entire training set, which might help to avoid local minima.