Exercise R.5.6
Gradient descent: starting points that end in a local minimum
Consider \(f(x)=x^3-x^2\) and provide the range of starting points \(x_0\) such that gradient descent will get stuck in a local minimum, assuming a small enough learning rate to guarantee convergence. What happens if \(x_0=0\)?
Hint: No need to compute any derivatives or iterations. Does the global minimum even exist? Use https://www.desmos.com/calculator for plotting. Also, any small learning rate will eventually reach the local minimum given a good starting condition.