Understand that gradient descent's learning rate controls a fundamental tradeoff: too small and convergence crawls, too large and the process overshoots and diverges, and for a well-behaved quadratic there's an exact optimal rate that reaches the minimum in a single step.
Loss: f(x) = (x − target)²Gradient: f'(x) = 2(x − target)Update rule: x_{k+1} = x_k − α·f'(x_k)Keep the default learning rate of 0.1 and run the full 20 iterations. Confirm x lands within 0.5 of the target (5).
Ready to check