1: \begin{abstract}

2: This paper investigates the global convergence of stepsized Newton methods for convex functions. We propose several simple stepsize schedules with fast global convergence guarantees, up to $\mathcal O\lr k^{-3} \rr$, nearly matching lower complexity bounds $\Omega \lr k^{-3.5} \rr$ of second-order methods. For cases with multiple plausible smoothness parameterizations or an unknown smoothness constant, we introduce a stepsize backtracking procedure that ensures convergence as if the optimal smoothness parameters were known.

3: 

4: \end{abstract}

5: