During the optimization process of the neural network, which one of the following is correct?

Single initialization weights in the neural network would be required to since the loss function is always convex

Different initialization weights in the neural network would be required to prevent the overfitting

None of the answers

Different initialization weights in the neural network would be required to overcome the issue of local minimum in the loss function

Different initialization weights in the neural network would be required only for regression problems not classification problems