Abstract Locally Checkable Labeling () problems are graph problems in which a solution is correct if it satisfies some given constraints in the local neighborhood of each node. Example problems in this class include maximal matching, maximal independent set, and colorings. A successful line of research has been studying the complexities of s on paths/cycles, trees, and general graphs, providing many interesting results for the model of distributed computing. In this work, we initiate the study of problems in the low-space Massively Parallel Computation () model. In particular, on forests, we provide a method that, given the complexity of an problem in the model, automatically provides an exponentially faster algorithm for the low-space setting that uses optimal global memory, that is, truly linear. While restricting to forests may seem to weaken the results, we emphasize that all known (conditional) lower bounds for the setting are obtained through lower bounds for problems in the distributed setting in tree-like networks (either trees or high-girth graphs), and hence the problems that we study are challenging already on trees. Moreover, our algorithms use optimal global memory, i.e., memory linear in the number of edges of the graph. In contrast, most of the state-of-the-art algorithms use more than linear global memory. Further, they typically start with a dense graph, sparsify it, and then solve the problem on the residual graph, exploiting the relative increase in global memory. On forests this is not possible, hence using optimal memory requires new solutions.

We give a randomized Δ-coloring algorithm in the LOCAL model that runs in poly log log n rounds, where n is the number of nodes of the input graph and Δ is its maximum degree. This means that randomized Δ-coloring is a rare distributed coloring problem with an upper and lower bound in the same ballpark, poly log log n, given the known Ω(logΔ logn) lower bound [Brandt et al., STOC '16].

We consider the problem of designing deterministic graph algorithms for the model of Massively Parallel Computation (MPC) that improve with the sparsity of the input graph, as measured by the standard notion of arboricity. For the problems of maximal independent set (MIS), maximal matching (MM), and vertex coloring, we improve the state of the art as follows. Let λ denote the arboricity of the n-node input graph with maximum degree Δ.

We present a deterministic (1+ε)-approximate maximum matching algorithm in poly(1/ε) passes in the semi-streaming model, solving the long-standing open problem of breaking the exponential barrier in the dependence on 1/ε. Our algorithm exponentially improves on the well-known randomized (1/ε)O(1/ε)-pass algorithm from the seminal work by McGregor [APPROX05], the recent deterministic algorithm by Tirodkar with the same pass complexity [FSTTCS18]. Up to polynomial factors in 1/ε, our work matches the state-of-the-art deterministic (logn / loglogn) · (1/ε)-pass algorithm by Ahn and Guha [TOPC18], that is allowed a dependence on the number of nodes n. Our result also makes progress on the Open Problem 60 at sublinear.info.

Locally Checkable Labeling (LCL) problems are graph problems in which a solution is correct if it satisfies some given constraints in the local neighborhood of each node. Example problems in this class include maximal matching, maximal independent set, and coloring problems. A successful line of research has been studying the complexities of LCL problems on paths/cycles, trees, and general graphs, providing many interesting results for the LOCAL model of distributed computing. In this work, we initiate the study of LCL problems in the low-space Massively Parallel Computation (MPC) model. In particular, on forests, we provide a method that, given the complexity of an LCL problem in the LOCAL model, automatically provides an exponentially faster algorithm for the low-space MPC setting that uses optimal global memory, that is, truly linear. While restricting to forests may seem to weaken the result, we emphasize that all known (conditional) lower bounds for the MPC setting are obtained by lifting lower bounds obtained in the distributed setting in tree-like networks (either forests or high girth graphs), and hence the problems that we study are challenging already on forests. Moreover, the most important technical feature of our algorithms is that they use optimal global memory, that is, memory linear in the number of edges of the graph. In contrast, most of the state-of-the-art algorithms use more than linear global memory. Further, they typically start with a dense graph, sparsify it, and then solve the problem on the residual graph, exploiting the relative increase in global memory. On forests, this is not possible, because the given graph is already as sparse as it can be, and using optimal memory requires new solutions.

A long line of research about connectivity in the Massively Parallel Computation model has culminated in the seminal works of Andoni et al. [FOCS'18] and Behnezhad et al. [FOCS'19]. They provide a randomized algorithm for low-space MPC with conjectured to be optimal round complexity O(log D + log log_{m/n} n) and O(m) space, for graphs on n vertices with m edges and diameter D. Surprisingly, a recent result of Coy and Czumaj [STOC'22] shows how to achieve the same deterministically. Unfortunately, however, their algorithm suffers from large local computation time. We present a deterministic connectivity algorithm that matches all the parameters of the randomized algorithm and, in addition, significantly reduces the local computation time to nearly linear. Our derandomization method is based on reducing the amount of randomness needed to allow for a simpler efficient search. While similar randomness reduction approaches have been used before, our result is not only strikingly simpler, but it is the first to have efficient local computation. This is why we believe it to serve as a starting point for the systematic development of computation-efficient derandomization approaches in low-memory MPC.

Metric clustering is a fundamental primitive in machine learning with several applications for mining massive datasets. An important example of metric clustering is the k-center problem. While this problem has been extensively studied in distributed settings, all previous algorithms use Ω(k) space per machine and Ω(n k) total work. In this paper, we develop the first highly scalable approximation algorithm for k-center clustering, with O~(n^ε) space per machine and O~(n^(1+ε)) total work, for arbitrary small constant ε. It produces an O(log log log n)-approximate solution with k(1+o(1)) centers in O(log log n) rounds of computation.

We provide a tight analysis that settles the round complexity of the well-studied parallel randomized greedy MIS algorithm, thus answering the main open question of Blelloch, Fineman, and Shun [SPAA’12]. The parallel/distributed randomized greedy Maximal Independent Set (MIS) algorithm works as follows. An order of the vertices is chosen uniformly at random. Then, in each round, all vertices that appear before their neighbors in the order are added to the independent set and removed from the graph along with their neighbors. The main question of interest is the number of rounds it takes until the graph is empty. This algorithm has been studied since 1987, initiated by Coppersmith, Raghavan, and Tompa [FOCS’87], and the previously best known bounds were O (log n ) rounds in expectation for Erdős-Rényi random graphs by Calkin and Frieze [Random Struc. Alg.’90] and O (log 2 n ) rounds with high probability for general graphs by Blelloch, Fineman, and Shun [SPAA’12]. We prove a high probability upper bound of O (log n ) on the round complexity of this algorithm in general graphs and that this bound is tight. This also shows that parallel randomized greedy MIS is as fast as the celebrated algorithm of Luby [STOC’85, JALG’86].

The Massively Parallel Computation (MPC) model serves as a common abstraction of many modern large-scale parallel computation frameworks and has recently gained a lot of importance, especially in the context of classic graph problems. In this work, we mainly consider maximal matching and maximal independent set problems in the MPC model.

In this paper, we present new randomized algorithms that improve the complexity of the classic (Δ+1)-coloring problem, and its generalization (Δ+1)-list-coloring, in three well-studied models of distributed, parallel, and centralized computation: Distributed Congested Clique: We present an O(1)-round randomized algorithm for (Δ + 1)-list-coloring in the congested clique model of distributed computing. This settles the asymptotic complexity of this problem. It moreover improves upon the O(log* Δ)-round randomized algorithms of Parter and Su [DISC'18] and O((log log Δ)⋅ log* Δ)-round randomized algorithm of Parter [ICALP'18].

We provide a tight analysis which settles the round complexity of the well-studied parallel randomized greedy MIS algorithm, thus answering the main open question of Blelloch, Fineman, and Shun [SPAA’12]. The parallel/distributed randomized greedy Maximal Independent Set (MIS) algorithm works as follows. An order of the vertices is chosen uniformly at random. Then, in each round, all vertices that appear before their neighbors in the order are added to the independent set and removed from the graph along with their neighbors. The main question of interest is the number of rounds it takes until the graph is empty. This algorithm has been studied since 1987, initiated by Coppersmith, Raghavan, and Tompa [FOCS’87], and the previously best known bounds were O(log n) rounds in expectation for Erdős-Rényi random graphs by Calkin and Frieze [Random Struc. & Alg. ’90] and O(log2 n) rounds with high probability for general graphs by Blelloch, Fineman, and Shun [SPAA’12]. We prove a high probability upper bound of O(log n) on the round complexity of this algorithm in general graphs, and that this bound is tight. This also shows that parallel randomized greedy MIS is as fast as the celebrated algorithm of Luby [STOC’85, JALG’86].

Sampling constitutes an important tool in a variety of areas: from machine learning and combinatorial optimization to computational physics and biology. A central class of sampling algorithms is the Markov Chain Monte Carlo method, based on the construction of a Markov chain with the desired sampling distribution as its stationary distribution. Many of the traditional Markov chains, such as the Glauber dynamics, do not scale well with increasing dimension. To address this shortcoming, we propose a simple local update rule based on the Glauber dynamics that leads to efficient parallel and distributed algorithms for sampling from Gibbs distributions. Concretely, we present a Markov chain that mixes in O(log n) rounds when Dobrushin's condition for the Gibbs distribution is satisfied. This improves over the LubyGlauber algorithm by Feng, Sun, and Yin [PODC'17], which needs O(Delta log n) rounds, and their LocalMetropolis algorithm, which converges in O(log n) rounds but requires a considerably stronger mixing condition. Here, n denotes the number of nodes in the graphical model inducing the Gibbs distribution, and Delta its maximum degree. In particular, our method can sample a uniform proper coloring with alpha Delta colors in O(log n) rounds for any alpha >2, which almost matches the threshold of the sequential Glauber dynamics and improves on the alpha>2 + sqrt{2} threshold of Feng et al.

We present a deterministic distributed algorithm that computes a (2Δ-1)-edge-coloring, or even list-edge-coloring, in any n-node graph with maximum degree Δ, in O(log8Δ·log n) rounds. This answers one of the long-standing open questions of distributed graph algorithms} from the late 1980s, which asked for a polylogarithmic-time algorithm. See, e.g., Open Problem 4 in the Distributed Graph Coloring book of Barenboim and Elkin. The previous best round complexities were 2O(√(log n)by Panconesi and Srinivasan [STOC'92] and Õ(√(Δ)) + O(log* n) by Fraigniaud, Heinrich, and Kosowski [FOCS'16]. A corollary of our deterministic list-edge-coloring also improves the randomized complexity of (2Δ-1)-edge-coloring to poly(log log n) rounds. The key technical ingredient is a deterministic distributed algorithm for hypergraph maximal matching, which we believe will be of interest beyond this result. In any hypergraph of rank r - where each hyperedge has at most r vertices - with n nodes and maximum degree Δ, this algorithm computes a maximal matching in O(r5log6+log rΔ·log n) rounds. This hypergraph matching algorithm and its extensions also lead to a number of other results. In particular, we obtain a polylogarithmic-time deterministic distributed maximal independent set (MIS) algorithm for graphs with bounded neighborhood independence, hence answering Open Problem 5 of Barenboim and Elkins book, a ((log Δ/ε)O(log 1/ε))-round deterministic algorithm for (1+ε)-approximation of maximum matching, and a quasi-polylogarithmic-time deterministic distributed algorithm for orienting λ-arboricity graphs with out-degree at most ⌈(1+ε)λ⌉, for any constant ε>0, hence partially answering Open Problem 10 of Barenboim and Elkin's book.

We present improved deterministic distributed algorithms for a number of well-studied matching problems, which are simpler, faster, more accurate, and/or more general than their known counterparts. The common denominator of these results is a deterministic distributed rounding method for certain linear programs, which is the first such rounding method, to our knowledge. A sampling of our end results is as follows. - An O(log^2 Delta log n)-round deterministic distributed algorithm for computing a maximal matching, in n-node graphs with maximum degree Delta. This is the first improvement in about 20 years over the celebrated O(log^4 n)-round algorithm of Hanckowiak, Karonski, and Panconesi [SODA'98, PODC'99]. - A deterministic distributed algorithm for computing a (2+epsilon)-approximation of maximum matching in O(log^2 Delta log(1/epsilon) + log^* n) rounds. This is exponentially faster than the classic O(Delta + log^* n)-round 2-approximation of Panconesi and Rizzi [DIST'01]. With some modifications, the algorithm can also find an epsilon-maximal matching which leaves only an epsilon-fraction of the edges on unmatched nodes. - An O(log^2 Delta log(1/epsilon) + log^* n)-round deterministic distributed algorithm for computing a (2+epsilon)-approximation of a maximum weighted matching, and also for the more general problem of maximum weighted b-matching. These improve over the O(log^4 n log_(1+epsilon) W)-round (6+epsilon)-approximation algorithm of Panconesi and Sozio [DIST'10], where W denotes the maximum normalized weight. - A deterministic local computation algorithm for a (2+epsilon)-approximation of maximum matching with 2^O(log^2 Delta) log^* n queries. This improves almost exponentially over the previous deterministic constant approximations which have query-complexity of 2^Omega(Delta log Delta) log^* n.