We provide faster running times for exactly solving discounted Markov Decision Processes (DMDPs) in strongly polynomial time. We obtain our results by efficiently reducing computing optimal values and policies in DMDPs to the easier tasks of policy evaluation and computing approximately optimal values. We provide both a straightforward deterministic reduction and a more efficient randomized variant that, together with advances in approximately solving DMDPs, yield our results.

In this paper we study the problem of minimizing a submodular function $f: 2^{V} \rightarrow \mathbb{R}$ that is guaranteed to have a k-sparse minimizer. We give a deterministic algorithm that computes an additive $\epsilon$-approximate minimizer of such f in $\widetilde{O}(\operatorname{poly}(k) \log (|f| / \epsilon))$ parallel depth using a polynomial number of queries to an evaluation oracle of f, where $|f|=\max_{S \subseteq V}|f(S)|$. Further, we give a randomized algorithm that computes an exact minimizer of f with high probability using $\widetilde{O}(|V| \cdot \operatorname{poly}(k))$ queries and polynomial time. When $k=\widetilde{O}(1)$, our algorithms use either nearly-constant parallel depth or a nearly-linear number of evaluation oracle queries. All previous algorithms for this problem either use $\Omega(|V|)$ parallel depth or $\Omega\left(|V|^{2}\right)$ queries. In contrast to state-of-the-art weakly-polynomial and strongly-polynomial time algorithms for SFM, our algorithms use first-order optimization methods, e.g., mirror descent and follow the regularized leader. We introduce what we call sparse dual certificates, which encode information on the structure of sparse minimizers, and both our parallel and sequential algorithms provide new algorithmic tools for allowing first-order optimization methods to efficiently compute them. Correspondingly, our algorithm does not invoke fast matrix multiplication or general linear system solvers and in this sense is more combinatorial than previous state-of-the-art methods.

We consider the parallel complexity of submodular function minimization (SFM). We provide a pair of methods which obtain two new query versus depth trade-offs a submodular function defined on subsets of $n$ elements that has integer values between $-M$ and $M$. The first method has depth $2$ and query complexity $n^{O(M)}$ and the second method has depth $\widetilde{O}(n^{1/3} M^{2/3})$ and query complexity $O(\mathrm{poly}(n, M))$. Despite a line of work on improved parallel lower bounds for SFM, prior to our work the only known algorithms for parallel SFM either followed from more general methods for sequential SFM or highly-parallel minimization of convex $\ell_2$-Lipschitz functions. Interestingly, to obtain our second result we provide the first highly-parallel algorithm for minimizing $\ell_\infty$-Lipschitz function over the hypercube which obtains near-optimal depth for obtaining constant accuracy.

We provide a generic technique for constructing families of submodular functions to obtain lower bounds for submodular function minimization (SFM). Applying this technique, we prove that any deterministic SFM algorithm on a ground set of n elements requires at least $\Omega(n\log n)$ queries to an evaluation oracle. This is the first super-linear query complexity lower bound for SFM and improves upon the previous best lower bound of 2n given by [Graur et al., ITCS 2020]. Using our construction, we also prove that any (possibly randomized) parallel SFM algorithm, which can make up to poly $(n)$ queries per round, requires at least $\Omega(n/\log n)$ rounds to minimize a submodular function. This improves upon the previous best lower bound of $\tilde{\Omega}(n^{1/3})$ rounds due to [Chakrabarty et al., FOCS 2021], and settles the parallel complexity of query-efficient SFM up to logarithmic factors due to a recent advance in [Jiang, SODA 2021].

We provide approximation algorithms for two problems, known as NECKLACE SPLITTING and $ε$-CONSENSUS SPLITTING. In the problem $ε$-CONSENSUS SPLITTING, there are $n$ non-atomic probability measures on the interval $[0, 1]$ and $k$ agents. The goal is to divide the interval, via at most $n (k-1)$ cuts, into pieces and distribute them to the $k$ agents in an approximately equitable way, so that the discrepancy between the shares of any two agents, according to each measure, is at most $2 ε/ k$. It is known that this is possible even for $ε= 0$. NECKLACE SPLITTING is a discrete version of $ε$-CONSENSUS SPLITTING. For $k = 2$ and some absolute positive constant $ε$, both of these problems are PPAD-hard. We consider two types of approximation. The first provides every agent a positive amount of measure of each type under the constraint of making at most $n (k - 1)$ cuts. The second obtains an approximately equitable split with as few cuts as possible. Apart from the offline model, we consider the online model as well, where the interval (or necklace) is presented as a stream, and decisions about cutting and distributing must be made on the spot. For the first type of approximation, we describe an efficient algorithm that gives every agent at least $\frac{1}{nk}$ of each measure and works even online. For the second type of approximation, we provide an efficient online algorithm that makes $\text{poly}(n, k, ε)$ cuts and an offline algorithm making $O(nk \log \frac{k}ε)$ cuts. We also establish lower bounds for the number of cuts required in the online model for both problems even for $k=2$ agents, showing that the number of cuts in our online algorithm is optimal up to a logarithmic factor.

We consider submodular function minimization in the oracle model: given black-box access to a submodular set function f:2^[n] → ℝ, find an element of arg min_S {f(S)} using as few queries to f(⋅) as possible. State-of-the-art algorithms succeed with Õ(n²) queries [Yin Tat Lee et al., 2015], yet the best-known lower bound has never been improved beyond n [Nicholas J. A. Harvey, 2008]. We provide a query lower bound of 2n for submodular function minimization, a 3n/2-2 query lower bound for the non-trivial minimizer of a symmetric submodular function, and a binom{n}{2} query lower bound for the non-trivial minimizer of an asymmetric submodular function. Our 3n/2-2 lower bound results from a connection between SFM lower bounds and a novel concept we term the cut dimension of a graph. Interestingly, this yields a 3n/2-2 cut-query lower bound for finding the global mincut in an undirected, weighted graph, but we also prove it cannot yield a lower bound better than n+1 for s-t mincut, even in a directed, weighted graph.