1: \begin{abstract}

2: We study the densest subgraph problem and give algorithms via multiplicative

3: weights update and area convexity that converge in $O\left(\frac{\log m}{\epsilon^{2}}\right)$

4: and $O\left(\frac{\log m}{\epsilon}\right)$ iterations, respectively,

5: both with nearly-linear time per iteration. Compared with the work

6: by \citet{bahmani2014efficient}, our MWU algorithm uses a very different

7: and much simpler procedure for recovering the dense subgraph from

8: the fractional solution and does not employ a binary search. Compared

9: with the work by \citet{boob2019faster}, our algorithm via area convexity

10: improves the iteration complexity by a factor $\Delta$---the maximum

11: degree in the graph, and matches the fastest theoretical runtime currently

12: known via flows \citep{chekuri2022densest} in total time. Next, we

13: study the dense subgraph decomposition problem and give the first

14: practical iterative algorithm with linear convergence rate $O\left(mn\log\frac{1}{\epsilon}\right)$

15: via accelerated random coordinate descent. This significantly improves

16: over $O\left(\frac{m\sqrt{mn\Delta}}{\epsilon}\right)$ time of the

17: FISTA-based algorithm by \citet{harb2022faster}. In the high precision

18: regime $\epsilon\ll\frac{1}{n}$ where we can even recover the exact

19: solution, our algorithm has a total runtime of $O\left(mn\log n\right)$,

20: matching the  exact algorithm via parametric flows \citep{gallo1989fast}.

21: Empirically, we show that this algorithm is very practical and scales

22: to very large graphs, and its performance is competitive with widely

23: used methods that have significantly weaker theoretical guarantees.

24: 

25: \end{abstract}

26: