English

Faster p-norm minimizing flows, via smoothed q-norm problems

Data Structures and Algorithms 2020-01-10 v2 Numerical Analysis Numerical Analysis Optimization and Control

Abstract

We present faster high-accuracy algorithms for computing p\ell_p-norm minimizing flows. On a graph with mm edges, our algorithm can compute a (1+1/poly(m))(1+1/\text{poly}(m))-approximate unweighted p\ell_p-norm minimizing flow with pm1+1p1+o(1)pm^{1+\frac{1}{p-1}+o(1)} operations, for any p2,p \ge 2, giving the best bound for all p5.24.p\gtrsim 5.24. Combined with the algorithm from the work of Adil et al. (SODA '19), we can now compute such flows for any 2pmo(1)2\le p\le m^{o(1)} in time at most O(m1.24).O(m^{1.24}). In comparison, the previous best running time was Ω(m1.33)\Omega(m^{1.33}) for large constant p.p. For pδ1logm,p\sim\delta^{-1}\log m, our algorithm computes a (1+δ)(1+\delta)-approximate maximum flow on undirected graphs using m1+o(1)δ1m^{1+o(1)}\delta^{-1} operations, matching the current best bound, albeit only for unit-capacity graphs. We also give an algorithm for solving general p\ell_{p}-norm regression problems for large p.p. Our algorithm makes pm13+o(1)log2(1/ε)pm^{\frac{1}{3}+o(1)}\log^2(1/\varepsilon) calls to a linear solver. This gives the first high-accuracy algorithm for computing weighted p\ell_{p}-norm minimizing flows that runs in time o(m1.5)o(m^{1.5}) for some p=mΩ(1).p=m^{\Omega(1)}. Our key technical contribution is to show that smoothed p\ell_p-norm problems introduced by Adil et al., are interreducible for different values of p.p. No such reduction is known for standard p\ell_p-norm problems.

Keywords

Cite

@article{arxiv.1910.10571,
  title  = {Faster p-norm minimizing flows, via smoothed q-norm problems},
  author = {Deeksha Adil and Sushant Sachdeva},
  journal= {arXiv preprint arXiv:1910.10571},
  year   = {2020}
}

Comments

ACM-SIAM Symposium on Discrete Algorithms (SODA 2020)

R2 v1 2026-06-23T11:52:37.736Z