What is initialization in Gradient Descent?

Initialization is the process of setting starting values for model parameters.

Gradient Descent begins from initial weights. Poor initialization can slow convergence or trap optimization in bad regions. Techniques like Xavier and He initialization improve performance in deep learning.