1: \begin{abstract}

2: We present an accelerated, or `look-ahead' version of the Newton--Dinkelbach method, a well-known technique for solving fractional and parametric optimization problems. 

3: This acceleration halves the Bregman divergence between the current iterate and the optimal solution within every two iterations. 

4: Using the Bregman divergence as a potential in conjunction with combinatorial arguments, we obtain strongly polynomial algorithms in three applications domains:

5: {\em (i)} For linear fractional combinatorial optimization, we show a convergence bound of $O(m\log m)$ iterations; the previous best bound was $O(m^2\log m)$ by Wang et al.~(2006).

6: {\em (ii)} We obtain a strongly polynomial label-correcting algorithm for solving  linear feasibility systems with two variables per inequality (2VPI).

7: For a 2VPI system with $n$ variables and $m$ constraints, our algorithm runs in $O(mn)$ iterations.

8: Every iteration takes $O(mn)$ time for general 2VPI systems, and 

9: $O(m + n\log n)$ time for the special case of deterministic Markov Decision Processes (DMDPs).

10: This extends and strengthens a previous result by Madani~(2002) that showed a weakly polynomial bound for a variant of the Newton–Dinkelbach

11: method for solving DMDPs.

12: {\em (iii)} We give a simplified variant of the parametric submodular function minimization result by Goemans et al.~(2017).

13: \end{abstract}

14: