An optimization algorithm used to minimize loss functions by iteratively adjusting model parameters in the direction of steepest descent. Gradient descent and its variants are fundamental to training neural networks, finding optimal parameter values through repeated updates based on error gradients.
