Search for: All records

We provide a general method to convert a "primal" black-box algorithm for solving regularized convex-concave minimax optimization problems into an algorithm for solving the associated dual maximin optimization problem. Our method adds recursive regularization over a logarithmic number of rounds where each round consists of an approximate regularized primal optimization followed by the computation of a dual best response. We apply this result to obtain new state-of-the-art runtimes for solving matrix games in specific parameter regimes, obtain improved query complexity for solving the dual of the CVaR distributionally robust optimization (DRO) problem, and recover the optimal query complexity for finding a stationary point of a convex function. 

The authors present an algorithm that, given an n-vertex m-edge Eulerian graph with polynomially bounded weights, computes an 𝑂 ~ ( 𝑛 log ⁡ 2 𝑛 ⋅ 𝜀 − 2 ) O ~ (nlog 2 n⋅ε −2 )-edge 𝜀 ε-approximate Eulerian sparsifier with high probability in 𝑂 ~ ( 𝑚 log ⁡ 3 𝑛 ) O ~ (mlog 3 n) time (where 𝑂 ~ ( ⋅ ) O ~ (⋅) hides polyloglog(n) factors). By a reduction from Peng-Song (STOC ’22), this yields an 𝑂 ~ ( 𝑚 log ⁡ 3 𝑛 + 𝑛 log ⁡ 6 𝑛 ) O ~ (mlog 3 n+nlog 6 n)-time algorithm for solving n-vertex m-edge Eulerian Laplacian systems with polynomially bounded weights with high probability, improving on the previous state-of-the-art runtime of Ω ( 𝑚 log ⁡ 8 𝑛 + 𝑛 log ⁡ 23 𝑛 ) Ω(mlog 8 n+nlog 23 n). They also provide a polynomial-time algorithm that computes sparsifiers with 𝑂 ( min ⁡ ( 𝑛 log ⁡ 𝑛 ⋅ 𝜀 − 2 + 𝑛 log ⁡ 5 / 3 𝑛 ⋅ 𝜀 − 4 / 3 , 𝑛 log ⁡ 3 / 2 𝑛 ⋅ 𝜀 − 2 ) ) O(min(nlogn⋅ε −2 +nlog 5/3 n⋅ε −4/3 ,nlog 3/2 n⋅ε −2 )) edges, improving the previous best bounds. Furthermore, they extend their techniques to yield the first 𝑂 ( 𝑚 ⋅ polylog ( 𝑛 ) ) O(m⋅polylog(n))-time algorithm for computing 𝑂 ( 𝑛 𝜀 − 1 ⋅ polylog ( 𝑛 ) ) O(nε −1 ⋅polylog(n))-edge graphical spectral sketches, along with a natural Eulerian generalization. Unlike prior approaches using short cycle or expander decompositions, their algorithms leverage a new effective resistance decomposition scheme, combined with natural sampling and electrical routing for degree balance. The analysis applies asymmetric variance bounds specialized to Eulerian Laplacians and tools from discrepancy theory. 

This paper presents a new parallel algorithm for minimizing Lipschitz convex functions using a stochastic subgradient oracle. The proposed method matches the state-of-the-art in terms of total queries and query depth (parallel rounds of queries) from [CJJLLST23], while improving the computational depth by a polynomial factor for sufficiently small accuracy. When combined with previous methods, this result closes the gap between the best-known query depth and computational depth in parallel stochastic convex optimization. The approach builds on the ball acceleration framework from [CJJJLST20, ACJJS21], which reduces optimization to minimizing a Gaussian-convolved regularization of the function within Euclidean balls. By developing new stability properties of the Hessian of this induced function, the authors reduce ball-constrained problems to stochastic unconstrained quadratic minimization. Although concentration results for the asymmetric Hessian approximations are lacking, the authors design an efficient parallel method for solving these quadratics. Interestingly, the algorithm can be further enhanced using fast matrix multiplication, yielding nearly-linear work if the matrix multiplication exponent is 2. 

In this paper we provide anO(mloglog^O(1)nlog (1/ϵ))-expected time algorithm for solving Laplacian systems onn-nodem-edge graphs, improving upon the previous best expected runtime of\(O(m \sqrt {\log n} \mathrm{log log}^{O(1)} n \log (1/\epsilon)) \)achieved by (Cohen, Kyng, Miller, Pachocki, Peng, Rao, Xu 2014). To obtain this result we provide efficient constructions of low spectral stretch graph approximations with improved stretch and sparsity bounds. As motivation for this work, we show that for every set of vectors in\(\mathbb {R}^d \)(not just those induced by graphs) and all integerk> 1 there exist an ultra-sparsifier withd− 1 +O(d/k) re-weighted vectors of relative condition number at mostk². For smallk, this improves upon the previous best known multiplicative factor of\(k \cdot \tilde{O}(\log d) \), which is only known for the graph case. Additionally, in the graph case we employ our low-stretch subgraph construction to obtainn− 1 +O(n/k)-edge ultrasparsifiers of relative condition numberk^{1 +o(1)}fork=ω(log ^δn) for anyδ> 0: this improves upon the previous work fork=o(exp (log ^{1/2 −δ}n)). 

We provide faster algorithms for approximating the optimal transport distance, e.g. earth mover's distance, between two discrete probability distributions on n elements. We present two algorithms that compute couplings between marginal distributions with an expected transportation cost that is within an additive ϵ of optimal in time O(n^2/eps); one algorithm is straightforward to parallelize and implementable in depth O(1/eps). Further, we show that additional improvements on our results must be coupled with breakthroughs in algorithmic graph theory.